Large language models

Canadian computer scientist whose work on sequence models, word embeddings and attention helped lay foundations for modern language models. He shared the 2018 ACM A.M. Turing Award with Geoffrey Hinton and Yann LeCun.

Microsoft is an American technology company founded in 1975 and headquartered in Redmond, Washington. Its AI work combines a partnership with OpenAI, the Copilot products and its own in-house MAI and Phi models.

Dario Amodei is the co-founder and CEO of Anthropic. He previously worked at OpenAI, where he was Vice President of Research and co-authored the GPT-2 paper and led the research for GPT-3.

Sam Altman is the co-founder and CEO of OpenAI. He ran the startup Loopt, was president of Y Combinator from 2014 to 2019, and was removed as OpenAI’s CEO in 2023 before returning days later.

Israeli-Canadian computer scientist born in 1986 who co-built AlexNet, co-founded OpenAI and served as its chief scientist, and later founded Safe Superintelligence Inc.

Published in 1997 by Sepp Hochreiter and Jürgen Schmidhuber, the LSTM paper introduced a memory cell with gates and a "constant error carousel" that lets recurrent networks learn across long time gaps. It shaped the recurrent models that came before Transformers.

ByteDance is a Chinese internet company founded in 2012, whose AI work includes the Doubao chatbot, the TRAE coding tool, the Seed model team and the Volcano Engine cloud platform.[^1]

Meta AI is the AI research and model-building organisation of Meta Platforms. Founded in late 2013 as Facebook AI Research, it produced PyTorch and the Llama and Muse model families, and it is separate from the Meta AI assistant product.

Attention lets a neural network weigh the parts of its input by relevance. Self-attention applies this idea inside a single sequence, so each token gathers information from every other token. It is the core operation of the Transformer.

Submitted to arXiv in September 2014 by Dzmitry Bahdanau, Kyunghyun Cho and Yoshua Bengio, this paper let a translation model look back at the whole source sentence while producing each word. This mechanism, called attention, later became a basis for Transformer models.

Ilya Sutskever, Oriol Vinyals and Quoc Le's 2014 paper showed that two LSTMs, one reading the input and one writing the output, could translate sentences and outperform a phrase-based statistical baseline. It set the encoder–decoder pattern behind later translation models.

OpenAI is a San Francisco AI research and deployment company founded in 2015 as a nonprofit. It makes the GPT model family and ChatGPT and filed confidentially for an IPO in June 2026.

NVIDIA's first data-center GPU built on its Volta architecture, announced in May 2017. It added Tensor Cores, dedicated matrix-multiply units; OpenAI later trained its GPT-3 models on V100s.

Reinforcement learning from human feedback trains a model toward what people prefer, using a reward model learned from human comparisons of its outputs. It became a standard step in building chat assistants.

A 2017 paper that introduced the Transformer, a neural network built entirely on attention with no recurrence. It set new machine translation results and became the architecture behind almost every large language model since.

The Transformer is a neural network architecture introduced in 2017 that relies on self-attention instead of recurrence or convolution. Most of today's large language models are built on it.

Pre-training trains a large model on huge amounts of unlabeled text, and fine-tuning then adapts it to a specific task with a small labeled dataset. This two-stage recipe is the basis of the term foundation model.

OpenAI's 2018 paper introducing GPT, a Transformer decoder pre-trained to predict the next token and then fine-tuned for specific tasks. It improved the state of the art on 9 of 12 tasks studied and set the recipe the GPT series followed.

BERT is a language model from Google that reads text in both directions and is pre-trained by filling in masked words. Fine-tuned per task, it set new records on question answering and language-understanding benchmarks and later went into Google Search.

Google's 2018 BERT paper introduced an encoder-only Transformer that reads text in both directions and is pre-trained by hiding words. It set new results on eleven language tasks and made download-and-fine-tune the default workflow.

Retrieval-augmented generation (RAG) lets a language model fetch relevant documents from an external store and put them into its prompt before answering, so it can use knowledge beyond its training data and cite where an answer came from.

A 2020 paper finding that language model loss falls predictably, as a power law, with model size, data and compute. It gave labs a quantitative reason to build much larger models and shaped the scale-up that followed.

Scaling laws are empirical trends showing that a language model's test loss falls smoothly, as a power law, when parameters, data and training compute grow. They help plan large training runs and explain the race for compute.

NVIDIA's A100 is a data-center GPU, the first built on the Ampere architecture. Announced on 14 May 2020, it introduced third-generation Tensor Cores, the TF32 format and Multi-Instance GPU, and became a standard building block for training large language models.

A 2020 paper that pairs a pre-trained seq2seq model with a dense Wikipedia index fetched by a neural retriever, setting state of the art on three open-domain QA tasks and producing more factual text than a parametric-only baseline.

GPT-3 is an autoregressive language model with 175 billion parameters, described in a May 2020 paper. It could perform new tasks from a few examples in the prompt, without any gradient updates, and it was first offered through OpenAI's API.

The GPT-3 paper showed that a language model with 175 billion parameters could perform many tasks from a few examples in its prompt, with no fine-tuning, and that this ability grew with model size.

A benchmark released in 2020 that tests language models across 57 subjects, from elementary mathematics to law, and showed that even the largest models of the time were far from expert-level.

Anthropic is an AI safety and research company founded in January 2021 by former OpenAI employees. It develops the Claude models and products, and its Responsible Scaling Policy sets out safety levels for its models.

A 2021 paper by Fedus, Zoph and Shazeer that simplifies Mixture-of-Experts routing so that language models can hold up to a trillion parameters while keeping per-example computation roughly constant.

A 2021 FAccT paper by Bender, Gebru, McMillan-Major and Shmitchell that asks whether language models can be too large and sets out recommendations on costs, data, research goals and alternative directions.

LoRA is a fine-tuning method that freezes a pre-trained model's weights and trains small low-rank matrices injected into each layer, cutting trainable parameters and memory by orders of magnitude.

A 2021 paper from OpenAI introduces Codex, a GPT model fine-tuned on public GitHub code, and HumanEval, an evaluation set that judges generated Python programs by whether they work rather than by how closely they match a reference.

A 2021 paper showing that finetuning a language model on many tasks described by instructions — instruction tuning — makes it much better at unseen tasks without examples, surpassing a larger model on most benchmarks tested.

InstructGPT is a family of GPT-3 models trained with reinforcement learning from human feedback so that they follow user instructions. OpenAI announced them in January 2022 and made them the default models in its API.

Chain of thought is the technique of having a language model write out its intermediate reasoning before answering. Reasoning models extend it by training the model with reinforcement learning to think at length before responding.

InstructGPT used human-written demonstrations, human rankings of outputs and reinforcement learning to make GPT-3 follow user instructions better. Labelers preferred its outputs even though it was far smaller than GPT-3.

The NVIDIA H100 is the first data-center GPU built on NVIDIA's Hopper architecture, announced in March 2022 with 80 billion transistors, HBM3 memory, FP8 Tensor Cores and a Transformer Engine aimed at large language models.

ReAct is a prompting method that has a language model interleave written reasoning with calls to outside tools or environments, improving question answering, fact verification and interactive decision-making.

An open-source framework released in October 2022 that lets developers connect large language models to data, APIs and other tools by composing them into chains and agents.

ChatGPT is OpenAI's AI chatbot, released as a research preview on 30 November 2022. It grew to hundreds of millions of weekly users and added agents, voice, browsing and workplace tools.

Google's consumer AI assistant, announced as Bard in 2023 and renamed Gemini in 2024. It is where ordinary users meet Google's Gemini models, first as a chat box and later through voice, research agents and subscriptions.

vLLM is an open-source serving system for large language models, built on PagedAttention, which manages the KV cache the way an operating system manages virtual memory. It reports 2-4x higher throughput than earlier serving systems at the same latency.

A family of foundation language models from Meta AI, released in 7B to 65B parameter sizes and trained on public data; its 13B model was reported to beat GPT-3 on most benchmarks.

LLaMA is a family of foundation language models from 7B to 65B parameters, trained only on public datasets; its 13B model beat GPT-3 on most benchmarks, and all models were released to researchers.

llama.cpp is an open-source inference engine written in C/C++ that runs large language models on consumer hardware. Created by Georgi Gerganov in March 2023, it helped make local model inference common and popularized the GGUF model file format.

GPT-4 is OpenAI's large multimodal model, released on 14 March 2023. It accepts image and text input, scored near the top of human test-takers on a simulated bar exam, and was succeeded by GPT-4 Turbo, GPT-4o and GPT-5.

OpenAI's technical report on GPT-4, submitted to arXiv in March 2023, describes a large multimodal Transformer model, reports human-level results on several professional benchmarks, and explains how the team made scaling behaviour predictable.

A crowdsourced benchmark in which anonymous pairs of language models battle and users vote, producing an Elo leaderboard; launched in 2023 and relaunched as LMArena in 2025.

A 2023 paper by Rafailov and colleagues that rewrites the reinforcement-learning-from-human-feedback objective so a language model can be aligned to preferences with a simple classification loss, without a separate reward model or reinforcement learning.

Function calling, now also called tool calling, lets an application describe external functions to a model, which then decides whether to emit JSON arguments for one of them — the route by which models reach data and actions outside their training data.

Claude apps (claude.ai) are Anthropic's chat assistant for consumers and businesses. The public beta opened with Claude 2 in 2023, and the apps have since added paid plans, Projects, web search and Research.

DeepSeek is a Chinese AI company founded in Hangzhou in July 2023 as a spin-off of the AGI research lab of High-Flyer, a hedge fund. It releases its main models with open weights.

A July 2023 paper from Meta led by Hugo Touvron that introduces and releases Llama 2, a family of pretrained and fine-tuned language models from 7 to 70 billion parameters, together with a detailed account of how the chat versions were tuned and made safer.

Mamba is a sequence-model architecture introduced by Albert Gu and Tri Dao that replaces attention with selective state space models, giving linear-time scaling, faster inference, and language-model quality competitive with Transformers twice its size.

Gemini 1.0 is the first generation of Google DeepMind's multimodal model family, announced on 6 December 2023 in Ultra, Pro and Nano sizes; Google reported its Ultra version as the first model to beat human experts on MMLU.

Mixtral 8x7B is a sparse mixture-of-experts language model from Mistral AI. A router sends each token to two of eight experts per layer, so the model holds 47B parameters but uses 13B during inference.

Claude 3 Opus, announced on 4 March 2024, was the most capable model in the Claude 3 family. It added vision and launched with a 200K-token context window and a Needle In A Haystack recall above 99%.

A report from Google's Gemini Team introducing the Gemini 1.5 family, multimodal models built around very long context, with Gemini 1.5 Pro recalling information across up to 10 million tokens.[^1]

Cerebras's third-generation wafer-scale AI chip, announced in March 2024: 4 trillion transistors on one 46,225-square-millimetre piece of silicon, built to keep very large models in a single memory space.

NVIDIA's GB200 NVL72 is a liquid-cooled rack that links 72 Blackwell GPUs and 36 Grace CPUs into a single NVLink domain aimed at trillion-parameter AI. Announced at GTC in March 2024, it is the building block of NVIDIA's Blackwell-generation DGX SuperPOD.

GPT-4o is a multimodal model from OpenAI, released in May 2024, that takes text, audio, image and video as input and generates text, audio and image output within one neural network. It was removed from ChatGPT in February 2026.

Claude 3.5 Sonnet, released on 20 June 2024, was the first model of the Claude 3.5 family. It ran twice as fast as Claude 3 Opus at the mid-tier price and powered the new Artifacts and Projects features.

Llama 3.1, released on 23 July 2024, added a 405-billion-parameter model to the Llama 3 family and extended its context to 128,000 tokens. Meta called the 405B the largest and most capable openly available foundation model.

Meta's 2024 paper describing Llama 3 as a family of foundation models — 8B, 70B and 405B parameters — trained on about 15 trillion tokens and reported to match leading models such as GPT-4 across a wide range of tasks.

OpenAI o1 is the first reasoning model series from OpenAI, introduced in September 2024. It thinks through a long internal chain of thought before answering, and it was the base for later reasoning models such as o3.

DeepSeek-V3 is an open-weight mixture-of-experts language model that DeepSeek released in December 2024. Each token uses 37 billion of its 671 billion parameters, and DeepSeek describes its chat model as comparable to leading closed-source systems.

A technical report from DeepSeek-AI on DeepSeek-V3, a mixture-of-experts language model with 671 billion total parameters, describing its architecture, training pipeline, and evaluations, including a full training run of 2.788 million H800 GPU hours.

DeepSeek-R1 is an open-weight reasoning model that DeepSeek released on 20 January 2025. Built on DeepSeek-V3-Base and trained with reinforcement learning, it is released under the MIT License, which allows commercial use and distillation.

The DeepSeek-R1 report described how a reasoning model was trained largely with reinforcement learning, rewarding correct answers, and reached performance comparable to OpenAI's o1 on math, code and reasoning benchmarks. Its weights were released openly.

Gemini 2.5 Pro is Google's first model in the Gemini 2.5 family, released in March 2025 as a "thinking model" with a 1-million-token context window, debuting at number one on LMArena.

OpenAI's second-generation reasoning model, announced in December 2024 and released in April 2025, which posted large gains over o1 on science, software-engineering and competitive-programming benchmarks before being succeeded by GPT-5.

Qwen3 is a family of large language models from Alibaba, released on 29 April 2025 under Apache 2.0. Its main feature is hybrid thinking, which lets one model reason step by step or answer directly.

The Qwen3 Technical Report presents Alibaba's Qwen3 family of dense and mixture-of-experts language models, which unify thinking and non-thinking modes, add a controllable thinking budget, and are released under the Apache 2.0 licence.

AMD's data-center GPU for AI, introduced in June 2025 on the CDNA 4 architecture, with 288 GB of HBM3E memory and very high low-precision throughput aimed at generative AI inference, positioned against Nvidia's Blackwell generation.

Kimi K2 is an open-weight mixture-of-experts language model from Moonshot AI, released in July 2025 with 1 trillion total parameters, 32 billion active per token and a 128K-token context.

The Gemini 2.5 technical report, submitted to arXiv on 7 July 2025, documents Google's Gemini 2.5 Pro and 2.5 Flash models: their reasoning, long-context video handling, and cost-versus-capability positioning.

Kimi K2 is Moonshot AI's trillion-parameter mixture-of-experts model, pre-trained with the MuonClip optimizer and post-trained on large-scale agentic data plus reinforcement learning; the report reports state-of-the-art results among open-source non-thinking models and releases base and post-trained checkpoints.

GPT-5 is OpenAI's fifth GPT foundation model, released on 7 August 2025. It pairs a fast model with a deeper reasoning model and a router that chooses between them for each request.

Gemini 3 Pro is the first model of Google's Gemini 3 generation, released in preview in November 2025 with a 1-million-token context window and top-ranked results on arena leaderboards and reasoning, coding and multimodal benchmarks.

NVIDIA Rubin is NVIDIA’s six-chip AI platform and the successor to Blackwell, announced at CES 2026. It targets lower-cost AI token generation, with partner products and cloud instances scheduled for 2026.

Zhipu's GLM-5 report describes a 744-billion-parameter mixture-of-experts model built on sparse attention and asynchronous reinforcement learning, aimed at moving AI-assisted programming from vibe coding toward agentic engineering.

Claude Mythos Preview, announced on 7 April 2026, is an unreleased frontier model that Anthropic made available only to restricted partners through Project Glasswing, a programme for securing critical software.

DeepSeek-V4, released in preview on 24 April 2026, comprises V4-Pro and V4-Flash, two open mixture-of-experts models with a 1M-token context. Its architecture targets long-context efficiency, and its later releases extended the series with agent upgrades and multimodal input.

A preview report on the DeepSeek-V4 series, describing a 1.6-trillion-parameter Pro model and a 284-billion-parameter Flash model, both with one-million-token context, plus the attention, connection and optimization changes that make that length affordable.

Claude Fable 5 is Anthropic's Mythos-class model for general use, released on 9 June 2026. Its sibling Claude Mythos 5 is the same model with some safeguards lifted for vetted users.

Kimi K3 is a 2.8-trillion-parameter open-weight multimodal model from Moonshot AI, released on 16 July 2026 with weights following on 27 July under the Kimi K3 License.

Kimi K3 is Moonshot AI's 2.8-trillion-parameter open mixture-of-experts model with native vision and a one-million-token context, released with full weights. Its report describes the architecture, training and serving work behind it, and reports frontier-level results that still trail the strongest proprietary models.

K2 Horizon is a family of six open language models from the Institute of Foundation Models, released in September 2026 with weights, code, training data and recipes under the Apache 2.0 licence.

GPT-6 is OpenAI's sixth-generation model family, launched from 3 September 2026 with GPT-6 Astra, followed by Sol, Luna and Instant. Astra was the first OpenAI model to reach its internal critical cybersecurity threshold.

Gemini 4 Argon is the first model in Google's Gemini 4 family, announced on 30 September 2026 and made generally available on 13 October 2026, built for long multi-step work and shipped with a one-million-token output limit.

Jais-3-375B-A23B is an open-weight Arabic-focused mixture-of-experts model from the Institute of Foundation Models: 375 billion parameters stored, 23 billion active per token, a 512,000-token context window and Apache 2.0 weights.

An AI agent is a system that perceives its environment and acts autonomously toward goals. The term covers everything from a thermostat to LLM-driven software that directs its own tool use.

Quantization stores a language model's weights and activations at lower numerical precision, cutting the memory and bandwidth that inference requires. Methods such as LLM.int8() and GPTQ made very large models runnable on far smaller hardware.

Hallucination is false or misleading information that a model presents as fact. It is a central failure mode of text-generating systems and a main obstacle to trusting their output.

Alignment is the problem of making a model's behaviour match what its users and designers actually intend, not just what its training objective literally rewards. Training on human preference comparisons is one practical route to it.

Taiwan-born computer scientist and executive who built research labs at Microsoft and Google in China, co-founded Sinovation Ventures, wrote widely read books on AI, and founded the foundation-model company 01.AI.

The Association for Computational Linguistics (ACL) is the main scientific and professional society for natural language processing. Founded in 1962, it runs one of the field's two leading conferences and publishes the journal Computational Linguistics.

Satya Nadella is the chief executive of Microsoft, who has led the company since 2014 and pushed it toward cloud computing and AI partnerships, including the 2019 deal with OpenAI.

Robin Li (李彦宏) is a Chinese technologist and businessman who co-founded Baidu in 2000 and has been its chief executive since 2004. He is best known for the company's search business and its ERNIE AI models.

Elon Musk is an entrepreneur who led SpaceX and Tesla, co-founded PayPal's predecessors and OpenAI, and founded xAI, the company behind the Grok models.

Czech computer scientist and lead author of word2vec, the 2013 methods that learn word vectors from text. His later work includes the fastText library and research on mathematical models that grow in complexity.

Mark Zuckerberg is the chairman and chief executive of Meta Platforms, the company formerly known as Facebook. He recruited the founding researchers of Facebook AI Research in 2013 and announced Meta Superintelligence Labs in 2025.

Mustafa Suleyman is a British AI entrepreneur who co-founded DeepMind and Inflection AI, and in 2024 became chief executive of Microsoft AI, leading its Copilot and in-house model work.

Liang Wenfeng is a Chinese entrepreneur who co-founded the quantitative hedge fund High-Flyer and founded the AI company DeepSeek in 2023. He is the corresponding author of the DeepSeek-R1 paper published in Nature in 2025.

A recurrent neural network (RNN) is a neural network with feedback connections that carry information from one step of a sequence to the next. Its long short-term memory variant (LSTM) was designed to keep useful memories over longer spans.

AI researcher and educator, born in Bratislava and raised partly in Toronto, who co-founded OpenAI, led AI at Tesla, taught Stanford's CS231n course, and later joined Anthropic's pre-training team.

Daniela Amodei is an American entrepreneur and co-founder of Anthropic, where she serves as President and leads teams focused on AI safety and alignment. She previously worked at Stripe and OpenAI.

Greg Brockman is a co-founder of OpenAI. He led early projects such as OpenAI Gym and OpenAI Five, later became president, and took over the product organisation in April 2026.

American AI researcher who co-founded OpenAI, led its reinforcement-learning team and helped develop the training methods behind ChatGPT; he later moved to Anthropic and then became chief scientist of Thinking Machines Lab.

Albanian-American engineer who served as OpenAI's chief technology officer, briefly as its interim CEO in November 2023, and later founded and leads Thinking Machines Lab.

Arthur Mensch is a French AI researcher who worked at DeepMind before co-founding Mistral AI in 2023, where he serves as chief executive. He is a co-author of the Chinchilla scaling paper and the Mistral 7B paper.

Tokenization splits text into the tokens a language model actually reads; most modern systems do it with byte pair encoding, a 1994 compression method adapted for translation in 2015.

Tencent is a Chinese internet company founded in 1998, known for QQ, WeChat and video games. Since 2023 its Hunyuan model line has been its main AI effort, later branded Hy.

iFlytek is a Chinese information technology company founded in 1999 and headquartered in Hefei. It is known for speech technology and for its Spark large language models, first released on 6 May 2023.

Google's research division, based in Mountain View and usually dated to 2000, produced the Transformer, BERT and T5 — the work underlying modern large language models.

Baidu is a Beijing-based Chinese internet company founded in 2000. It runs China's dominant search engine and develops the ERNIE model family and the ERNIE Bot chatbot.

Published in 2003 by Yoshua Bengio and colleagues, this paper proposed learning a vector for every word together with a neural network that predicts the next word. It introduced the core idea behind word embeddings and reached lower test perplexity than n-gram models.

A word embedding represents each word as a list of numbers learned from text, so that words used in similar ways get similar lists. Embeddings are a basic building block of modern language models.

Alibaba Cloud is the cloud computing arm of Alibaba Group and the developer of the Qwen (Tongyi Qianwen) large language model family, which is released as both open weights and proprietary models.

Databricks is a San Francisco software company founded in 2013 by seven people behind Apache Spark, which grew out of the University of California, Berkeley AMPLab project [1][2]. Its valuation rose from $38 billion in 2021 to $188 billion in 2026 [4][11].

Submitted to arXiv in January 2013 by Tomas Mikolov and colleagues at Google, this paper introduced CBOW and Skip-gram, two simple architectures that learn word vectors quickly from very large text collections. They are the basis of the word2vec project.

Nonprofit AI research institute founded by Paul Allen in 2014 and based in Seattle, known for Semantic Scholar and for the openly released OLMo family of language models, whose training data, code and checkpoints are published alongside the weights.

SenseTime is a Chinese AI company founded in October 2014 and listed in Hong Kong at the end of 2021. It develops the SenseNova large models and, in 2026, reported its first IFRS net profit since listing.

SQuAD is a reading-comprehension dataset of more than 100,000 crowdworker questions on Wikipedia articles, each answered by a span of its passage. Released in 2016, it gave machine comprehension a large, freely available benchmark with a measured human ceiling.

A mixture of experts model contains many sub-networks but routes each token to only a few of them. This lets total model size grow while the compute spent on each token stays much smaller than the total.

Tensor Cores are specialised units inside NVIDIA GPUs that perform a small matrix multiply-and-add in one step, making the matrix arithmetic behind neural-network training and inference dramatically faster.

A 2017 paper showing that an AI agent can learn a task from short human comparisons instead of a hand-written reward. The method, reinforcement learning from human feedback, later became central to training chat assistants.

AI21 Labs is an Israeli AI company founded in Tel Aviv in November 2017 that builds large language models, including the Jurassic and Jamba families, and the Wordtune writing assistant.

GLUE is a benchmark and analysis platform that evaluates natural language understanding models across a diverse set of existing tasks, with an online comparison platform and a hand-crafted diagnostic suite for linguistic analysis.

GPT-1 was OpenAI's first generative pre-trained transformer, published in June 2018. It showed that one language model, pre-trained on unlabelled text and then fine-tuned, could improve results on many language tasks.

Zhipu AI, also known as Z.ai, is a Beijing company founded in 2019 as a Tsinghua University spin-out. It develops the GLM model family and related chatbot, video, agent and coding products, and listed in Hong Kong in January 2026.

Cohere is a Toronto-based enterprise AI company founded in 2019. It builds the Command family of language models, sells them to businesses, and in September 2026 signed a definitive agreement to combine with Aleph Alpha.

MBZUAI, the Mohamed bin Zayed University of Artificial Intelligence, is a graduate AI university in Abu Dhabi, founded in 2019. It hosts the Institute of Foundation Models, which builds the K2, Jais and PAN model lines.

GPT-2 is a large language model that OpenAI announced in February 2019. Its largest version has 1.5 billion parameters. OpenAI first withheld the full model, then released it in stages through November 2019.

OpenAI's 2019 paper on GPT-2, a 1.5-billion-parameter Transformer trained on web text. It showed that a model trained only to predict the next word could perform several tasks when prompted, and its staged release sparked public debate.

Baidu's first ERNIE pre-trained language model, released in March 2019, added entity- and phrase-level masking to BERT-style pre-training and reported state-of-the-art results on five Chinese NLP tasks.

SuperGLUE is a benchmark for general-purpose language understanding, released in 2019 as a harder successor to GLUE. It bundles eight tasks, a single-number metric, an analysis toolkit and a public leaderboard.

XLNet is a generalized autoregressive pretraining method from a 2019 paper. It learns bidirectional contexts by maximizing expected likelihood over all factorization-order permutations and outperforms BERT on 20 tasks under comparable settings.

A replication study of BERT pretraining that measures how much key hyperparameters and training data size matter, finds BERT was significantly undertrained, and shows a properly trained BERT matches or beats later models, reaching state of the art on GLUE, RACE and SQuAD.

Huawei's Ascend 910 is an AI training processor launched in 2019 alongside the MindSpore framework, rated at 256 teraflops FP16 and used to train the PanGu-α language models on clusters of thousands of chips.

T5 is a family of Google Transformer models that recasts every language task as text in, text out. Its 11-billion-parameter version set state of the art on several benchmarks, and the code, models and C4 dataset were released openly.

A 2019 paper that casts every language task as text-to-text, systematically compares transfer-learning choices, releases the C4 corpus and models, and reports state-of-the-art results on summarization, question answering, and classification.

Google's TPU v4 is a machine-learning chip and supercomputer deployed since 2020, built into 4,096-chip pods with optically reconfigurable networking; PaLM and Gemini models were trained on it.

A 2020 paper by Clark, Luong, Le and Manning that pre-trains text encoders by detecting replaced tokens instead of filling in masked ones, matching much larger baselines with far less compute.

In-context learning is a large language model's ability to perform a new task from instructions and a few examples placed in the prompt, with no weight updates. Named and studied in the GPT-3 paper, it moved adaptation from fine-tuning to prompt writing.

GShard, submitted to arXiv in June 2020, is a set of annotation APIs and an XLA compiler extension from Google that lets existing model code express parallel computation, scaling a sparsely gated mixture-of-experts translation model past 600 billion parameters.

EleutherAI is a grassroots AI research collective that began as a Discord server in July 2020. It released The Pile and the open GPT-style models GPT-Neo, GPT-J and GPT-NeoX-20B, and in 2023 it became a nonprofit research institute.

The Pile is an 825 GiB English text dataset of 22 combined sources, released by EleutherAI on 31 December 2020 with an arXiv paper. It became the training data for several of the group's open models.

MATH is a dataset of 12,500 challenging competition mathematics problems, each with a full step-by-step solution. Released in 2021, it was built to measure and train mathematical reasoning in models.

LaMDA is a family of conversational language models that Google announced in 2021 and described in a 2022 paper, notable for its work on safety and factual grounding in dialogue and for later powering Bard.

GPT-J-6B is a six-billion-parameter open language model released by EleutherAI in June 2021 under the Apache 2.0 license. Its published scores sit close to GPT-3 Curie, making it a widely used free baseline for large-model research.

LoRA (low-rank adaptation) is a fine-tuning technique that freezes a pre-trained model's weights and trains small low-rank matrices instead, cutting the number of trainable parameters and the memory needed while keeping quality close to full fine-tuning.

ERNIE 3.0 is a 10-billion-parameter pre-trained language model from 2021 that trains on plain text plus a large-scale knowledge graph and fuses auto-encoding and auto-regressive networks, so one model handles both language understanding and generation.

OpenAI Codex is a 2021 GPT language model fine-tuned on public code from GitHub. It translates natural language into code, and was introduced alongside the HumanEval benchmark for testing code correctness.

A 2021 arXiv paper by Austin, Odena, Nye and eight other authors that introduced MBPP, a benchmark of 974 short Python programming tasks, and measured how language models from 244M to 137B parameters turn natural-language descriptions into working code.

Instruction tuning is a supervised fine-tuning stage that trains a large language model to follow user instructions; the FLAN paper showed it substantially improves zero-shot performance on unseen tasks.

GSM8K is an OpenAI benchmark of 8.5K grade school math word problems, released with a paper showing that even the largest Transformer models scored poorly, and that training verifiers to select among many candidate solutions improved results.

MiniMax is a Shanghai-based AI company founded in December 2021 by computer-vision researchers from SenseTime. It builds text, speech and video models and consumer apps such as Talkie and Hailuo AI, and listed in Hong Kong in January 2026.

A 2022 paper by Jason Wei and colleagues at Google Research showing that prompting a large language model with examples that include intermediate reasoning steps, rather than answers alone, substantially improves its performance on arithmetic, commonsense and symbolic reasoning tasks.

GPT-NeoX-20B is a 20-billion-parameter open autoregressive language model released by EleutherAI in February 2022, with weights under Apache 2.0 and its training code, described by its makers as the largest publicly accessible general-purpose model of its kind at the time.

GPT-3.5 is the family of OpenAI language models between GPT-3 and GPT-4, released through API models such as text-davinci-002 and text-davinci-003 and underlying the first version of ChatGPT.

NVIDIA's fourth-generation DGX system, announced in March 2022, packs eight H100 GPUs linked by NVLink plus BlueField-3 networking, and is rated at 32 petaflops of FP8 AI performance for enterprise AI infrastructure.

A DeepMind study showing that, for a fixed computing budget, language models should be trained on far more data than was then standard; its 70-billion-parameter Chinchilla beat larger contemporaries.

PaLM is a 540-billion-parameter dense Transformer language model announced by Google Research in April 2022, trained with the Pathways system across 6,144 TPU v4 chips; it set strong few-shot results on many benchmarks.

BIG-bench is a large community-built benchmark of 204 tasks designed to test language models on problems believed to lie beyond their current abilities, described in a paper posted in June 2022.

A 2022 paper arguing that some abilities of large language models appear only once models are large enough, and cannot be predicted by extrapolating from smaller models — so further scaling may keep widening what these models can do.

BLOOM is a 176-billion-parameter open multilingual language model released on 12 July 2022 by the BigScience project, with weights available under a responsible-use license.

Perplexity AI is a San Francisco company founded on 22 August 2022 by Aravind Srinivas and others. It runs an AI answer engine that cites web sources and has since added models, a browser and agents.

GLM-130B is a 130-billion-parameter open bilingual Chinese–English dense language model from Tsinghua University's KEG group and Zhipu AI, built to give researchers an open model that matches GPT-3 at the hundred-billion scale.

LlamaIndex is an open-source Python library for building retrieval-augmented generation applications and AI agents on top of large language models. It began in late 2022 as one engineer's small indexing utility, GPT Tree Index.

A 2022 paper by Percy Liang, Rishi Bommasani, Tony Lee and many co-authors that ran 30 language models through the same 42 scenarios and seven metrics, turning scattered benchmarks into one standardized evaluation.

Perplexity is an AI answer engine launched on 7 December 2022. It answers questions in natural language and cites its web sources, and it later added Deep Research, finance and shopping features and agents.

Constitutional AI is a training method introduced by Anthropic in December 2022, in which a model's behaviour is shaped by a written constitution of principles rather than by human labels of harmful outputs.

Constitutional AI, a 2022 paper from Anthropic, proposes training a harmless AI assistant with AI-generated feedback guided by a written set of principles, instead of human labels of harmful outputs.

An open bilingual chat model from Zhipu AI and Tsinghua KEG with about 6.2 billion parameters, small enough to run on one consumer GPU, with weights fully open for research.

A public Hugging Face leaderboard that ran a fixed set of benchmarks on open language models in one identical setup, from 2023 to 2025, so their scores could be reproduced and compared.

OpenRouter is a marketplace and gateway for large language models, started in early 2023, that lets developers reach hundreds of models from many providers through one interface. It reports more than 10 million users and a $113 million Series B.

Direct preference optimization (DPO) is a method for fine-tuning language models on human preference data with a supervised-learning-style objective, removing the separate reward model and the reinforcement-learning loop of standard RLHF.

Prompt injection is an attack in which text inside a model's input overrides its original instructions, letting an adversary hijack a language-model application — including, as researchers showed in 2023, through content the application retrieves on its own.

Microsoft Copilot is the company's AI chatbot and assistant, which started in 2023 as Bing Chat and became a single consumer and business AI brand that now includes in-house models.

A 2023 paper by Schick, Dwivedi-Yu, Dessì and co-authors showing that a language model can teach itself when to call a calculator, search engine, Q&A system, translation system or calendar, and how to use the results.

Chinese AI company founded in Beijing in 2023 by Tsinghua schoolfriends Yang Zhilin, Zhou Xinyu and Wu Yuxin; maker of the long-context Kimi assistant and the K-series models, and one of China's fastest-rising model developers by valuation.

The H800 is a China-market version of NVIDIA's H100 data-center GPU, adapted to fit the US export rules of October 2022 by cutting chip-to-chip transfer speed. It was later used to train DeepSeek-V3 and was covered by tighter US controls in October 2023.

01.AI is a Beijing-based artificial intelligence company founded in 2023 by Kai-Fu Lee. It builds open and closed large language models, including the Yi series, and the Wanzhi enterprise platform.

GPT-3.5 Turbo was the chat-optimised GPT-3.5 model OpenAI released through its API on 1 March 2023, priced at a tenth of the cost of earlier models and later replaced in ChatGPT by GPT-4o mini.

xAI is an AI company founded by Elon Musk that builds the Grok models and chatbot. It was acquired by SpaceX in 2026 and has since been renamed SpaceXAI and then SpaceXSI.

Claude is the AI assistant Anthropic announced on 14 March 2023, its first publicly available assistant. It was built around training AI to be helpful, honest and harmless, and it started the Claude model line.

Microsoft 365 Copilot is Microsoft's AI assistant for Word, Excel, PowerPoint, Outlook and Teams, which pairs large language models with a customer's business data.

ERNIE Bot (文心一言) is Baidu's chatbot, first released to invited testers in March 2023 and opened to the public that August. It has been offered in free and paid tiers and renamed Wenxiaoyan in 2024.

A March 2023 paper by Microsoft researchers testing an early GPT-4, arguing it shows strikingly general, near-human performance across many domains and could be seen as an early but incomplete step toward AGI.

An open-source autonomous agent released in March 2023 that uses OpenAI's large language models, such as GPT-4, to break a user's goal into sub-tasks and carry them out with tools like web browsing.

Mistral AI is a Paris-based company founded in April 2023. It builds open-weight and commercial large language models, the Le Chat assistant and the Vibe coding agent, and it has grown quickly through several large funding rounds.

The Qwen app is Alibaba's consumer chatbot built on its Qwen models. It began as a beta of Tongyi Qianwen in April 2023 and has since added app integrations and reached 234 million users by May 2026.

StepFun (阶跃星辰) is a Shanghai AI company founded in April 2023 by former Microsoft staff. It builds multimodal Step-series models and has open-sourced several of them.

Baichuan Intelligence (百川智能) is a Beijing AI company founded in April 2023 by Wang Xiaochuan. It released open-source language models from 2023 and later focused on applying AI to healthcare.

SenseNova is SenseTime's family of foundation models, unveiled on 10 April 2023 with a chatbot, an image generator and an avatar video generator, first offered to government and corporate clients rather than the public.

SparkDesk is iFlytek's cognitive large model, unveiled on 6 May 2023 as the company's answer to ChatGPT in education and enterprise uses, built on Huawei Ascend chips and followed by rapid version updates through 2024.

PaLM 2 is Google's second-generation Pathways language model, announced on 10 May 2023 in four sizes and used across more than 25 Google products, including Bard.

QLoRA is a finetuning method that trains a 65B-parameter language model on a single 48GB GPU while matching full 16-bit finetuning quality, using 4-bit quantization, double quantization and paged optimizers.

Sakana AI is a Tokyo research and development company founded in July 2023 by David Ha, Ren Ito and Llion Jones, known for evolutionary model merging, projects such as The AI Scientist, and Japanese-language models named Namazu.

Ollama is an open-source tool written in Go that lets people pull and run large language models locally on macOS, Linux and Windows, from a command line interface or a local REST API.

Claude 2 is the second major version of Anthropic's assistant model, released on 11 July 2023 with better performance and longer responses. It was the first Claude model open to the general public.

A Google research and note-taking tool that answers only from the sources a user supplies and cites the passages it used. Launched in 2023 under the name Project Tailwind, it became widely known for turning documents into podcast-style Audio Overviews.

Meta's family of open-weight language models released in July 2023 in 7B, 13B and 70B sizes, each with a dialogue-tuned Chat variant, issued under a custom licence that allowed free research and commercial use.

GGUF is a binary file format introduced in 2023 by the llama.cpp project that packs model tensors and metadata into a single file, and is widely used for distributing quantized language models for local inference.

Doubao is ByteDance's AI chatbot, opened for invitation testing in August 2023 and later offering text, image and voice generation, AI search and video generation.[^1][^2]

Qwen, known in Chinese as Tongyi Qianwen (通义千问), is Alibaba's family of large language models. Its first open-weight model, Qwen-7B, was released on 3 August 2023, and the family has grown through many later versions.

AutoGen is a Microsoft framework for building AI agents and applications, in which customizable, conversable agents combine large language models, human input and tools. It was introduced in an arXiv paper submitted on 16 August 2023.

Code Llama is a family of coding models that Meta released in August 2023 by further training Llama 2 on code. It ranges from 7 billion to 70 billion parameters and uses the Llama 2 community license.

Baichuan 2 is the second-generation open model family from Baichuan, released on 6 September 2023 in 7B and 13B sizes. Commercial use required an application and was limited by user numbers.

Hunyuan is Tencent's first proprietary large foundation model, unveiled in September 2023 and offered to enterprises through Tencent Cloud. The line was later branded Hy.

GPT-4V is a version of GPT-4 that accepts image input alongside text, documented in OpenAI's system card of 25 September 2023 and released to developers through preview API models; image understanding later became native in GPT-4o.

Mistral AI's first model: a 7.3-billion-parameter language model released in September 2023 under the Apache 2.0 license, built to match much larger models on standard benchmarks.

Meta AI is Meta's conversational assistant, introduced in US beta in September 2023 on WhatsApp, Messenger and Instagram. It was rebuilt on Llama 3 in 2024 and on Muse Spark in 2026.

Kimi is an AI chatbot from Moonshot AI, released in October 2023 and opened to the public the following month. It is known for very long context windows and has since added search, agent and coding features.

DSPy is a programming model that represents language-model pipelines as graphs of declarative modules and uses a compiler to optimize them against a chosen metric.

ERNIE 4.0 is Baidu's foundation model unveiled in October 2023, pitched as its most capable model and, in Robin Li's words, "not inferior in any aspect to GPT-4". It anchors Baidu's ERNIE Bot assistant and Qianfan platform.

TensorRT-LLM is NVIDIA's open-source library for speeding up and optimizing large language model inference on NVIDIA GPUs, publicly released on 19 October 2023 and usable free from GitHub or as part of NVIDIA NeMo.

CrewAI is an open-source Python framework for building groups of collaborating AI agents, created by João Moura in late 2023 and backed by $18 million in funding.

Yi-34B and Yi-6B are open bilingual base models from 01.AI, released in November 2023 and trained on 3 trillion tokens. Yi-34B ranked first among open pre-trained models at launch.

Grok is the chatbot from xAI, first offered in November 2023 to a small waitlist of US users. It has grown into a family of apps and models covering search, images, video and coding agents.

Grok-1 is xAI's first large language model, announced on 3 November 2023 as the engine of the Grok chatbot. Its 314-billion-parameter weights were released openly in March 2024.

GPT Actions let a custom GPT call external APIs, turning a user's request into the JSON input an API needs. Introduced by OpenAI on 6 November 2023, they grew out of the plugins beta and rely on OpenAPI schemas plus authentication settings.

GPT-4 Turbo is a cheaper, more capable version of OpenAI's GPT-4, announced on 6 November 2023 with a 128,000-token context window and lower API prices.

NVIDIA's H200, announced in November 2023, is a data center GPU that was the first to use HBM3e memory, offering 141 GB at 4.8 TB/s — nearly double the A100's capacity — and roughly double the H100's inference speed on Llama 2 70B.

Azure Maia 100 is Microsoft's first custom AI chip, announced in November 2023 and built to train and run large language models inside its own Azure data centres rather than being sold.

Claude 2.1, released on 21 November 2023, doubled the context window to 200,000 tokens and cut false statements by half compared with Claude 2.0. It added tool use and system prompts.

AMD's data-center GPU for generative AI, released in December 2023. Built on the CDNA 3 architecture with a chiplet design, it leads with 192 GB of HBM3 memory, which AMD contrasted with the 80 GB on Nvidia's H100.

An open-weight sparse mixture-of-experts language model from Mistral AI, released in December 2023: 46.7B total parameters but 12.9B used per token, a 32k-token context, and results above Llama 2 70B on most benchmarks.

SGLang is an open-source system for running complex language model programs: a frontend language plus a runtime that reuses cached attention state and speeds up structured output, reporting up to 6.4x higher throughput than prior inference systems.

OLMo is the Allen Institute for AI's open language model, released on 1 February 2024 in 1B and 7B sizes with its Dolma training data, training code, logs and checkpoints, all under Apache 2.0.

A mid-size multimodal model in Google's Gemini 1.5 generation, announced in February 2024, known for a context window that reached one million tokens and later two million.

Gemma is a family of lightweight open language models from Google DeepMind, released in February 2024 in 2B and 7B sizes, built from the same research and technology as Gemini and offered for commercial use under Google's terms.

Le Chat is Mistral AI's assistant, launched in beta on 26 February 2024. It gained web search, a Pro tier, Deep Research and other features over 2024 and 2025, and on 28 May 2026 it was merged into the Vibe agent.

Mistral Large is the French company Mistral AI's first flagship proprietary large language model, released on 26 February 2024 with a 32K-token context window, native fluency in five languages, and built-in function calling and JSON output.

Claude 3 Sonnet was the middle model of Anthropic's Claude 3 family, announced on 4 March 2024. It powered the free tier of claude.ai, offered a 200K-token context window, and was retired on 21 July 2025.

A benchmark for code LLMs that keeps collecting fresh contest problems from LeetCode, AtCoder and CodeForces, so that scores reflect ability rather than memorised test data, and that also tests code repair and execution reasoning.

Claude 3 Haiku is the smallest and fastest model in Anthropic's Claude 3 family, announced in March 2024 with low per-token pricing, a 200K-token context window and image input; retired in April 2026.

NVIDIA's data-center GPU for the Blackwell generation, announced in March 2024 as the successor to Hopper, built from two joined dies and paired with NVLink and FP4 support for large-scale AI training and inference.

Microsoft AI is the consumer AI organisation Microsoft formed in March 2024 to advance Copilot and its other AI products and research, led by Mustafa Suleyman. It develops the MAI family of in-house models.

EvoLLM-JP is a Japanese model from Sakana AI built by evolutionarily searching merges of three open-source models instead of training from scratch; the 7B result has mathematics ability none of its sources had alone.

DBRX is Databricks' openly licensed large language model, a mixture-of-experts transformer with 132 billion total parameters and 36 billion active per token, announced on 27 March 2024.

Jamba, released by AI21 Labs on 28 March 2024, is an open-weights model that combines Mamba state-space layers with Transformer attention. It has 52 billion total parameters and a 256k-token context window.

Cohere's 104-billion-parameter language model, announced in April 2024, is tuned for retrieval-augmented generation and multi-step tool use in business settings; it launched first on Microsoft Azure and was also released as open weights.

Llama 3 is an open model family from Meta, released on 18 April 2024 at 8 billion and 70 billion parameters. It was pretrained on over 15 trillion tokens of publicly available data.

The Meta Llama 3 Community License is the custom agreement Meta attaches to its Llama 3 models: free to use and modify, but with use restrictions, attribution duties and a user-scale threshold that keep it from being open source in the usual sense.

Phi-3 is Microsoft's family of small language models, launched in April 2024 with Phi-3-mini, a 3.8-billion-parameter model small enough to run on a phone.

SenseNova 5.0 is SenseTime's large language model released in April 2024, built on a mixture-of-experts architecture with a roughly 200K-token context window, which the company positioned as matching or surpassing GPT-4 Turbo.

Snowflake's open-weight model with 480 billion total and 17 billion active parameters, built as a dense-MoE hybrid, released in April 2024 under Apache 2.0 and trained for under 2 million dollars.

DeepSeek-V2 is a mixture-of-experts language model released in May 2024 with 236B total parameters, of which 21B are activated per token. It introduced Multi-head Latent Attention and used DeepSeekMoE for economical sparse training.

Yi-Large is a closed large language model with about one hundred billion parameters, released by 01.AI on 13 May 2024 and offered through its API platform; 01.AI reported top rankings for it on several public benchmarks.

AI Overviews are AI-written summaries placed at the top of Google Search results, launched across the US in May 2024 and later joined by AI Mode, a separate search experiment. They became one of the most widely used consumer applications of generative AI.

ByteDance's Doubao model family, launched through Volcano Engine in May 2024, offered general-purpose, speech and image models at prices far below rivals, helping set off a price war among Chinese providers.

Tencent Yuanbao is an AI assistant app from Tencent, launched on 30 May 2024, that runs on its Hunyuan models and later integrated DeepSeek-R1, HY 2.0 and Hy3.

MMLU-Pro is a 2024 benchmark that makes the widely used MMLU knowledge test harder: its questions are reasoning-focused and each has ten answer options instead of four.

Qwen2 is the Qwen family's June 2024 generation, with five sizes from 0.5B to 72B parameters, context lengths up to 128K tokens, and most sizes released under Apache 2.0.

Apple Intelligence is Apple's personal-intelligence system, announced at WWDC on 10 June 2024 for iOS 18, iPadOS 18 and macOS Sequoia. It runs models on device and uses Private Cloud Compute for larger requests.

τ-bench is a benchmark, submitted to arXiv in June 2024, that tests language agents in dynamic conversations with a simulated user, with domain-specific API tools and policy rules, and scores them by database state rather than by the text of their replies.

Safe Superintelligence Inc. (SSI) is an American AI company founded in June 2024 by Ilya Sutskever, Daniel Gross and Daniel Levy to build a safe superintelligence as its only product, without releasing anything in between.

FineWeb is a 15-trillion-token pretraining dataset built from 96 Common Crawl snapshots and released in 2024 with its curation code and ablation models; its educational subset, FineWeb-Edu, improves results on knowledge- and reasoning-heavy benchmarks.

iFlytek's large language model released on 27 June 2024 and trained entirely on a domestic platform co-built with Huawei; iFlytek presented it as reaching GPT-4 level while acknowledging weaker coding and multimodal abilities.

Step-2 is a trillion-parameter language model from StepFun built on a Mixture-of-Experts architecture. A preview appeared in March 2024; the official version launched at WAIC in July 2024.

A small, low-cost language model OpenAI released on 18 July 2024, priced far below GPT-4o and reported at 82% on MMLU; it replaced GPT-3.5 Turbo in ChatGPT on launch day.

Mistral AI's second-generation flagship model, released in July 2024 with 123 billion parameters, a 128K-token context window, wide language and code coverage, and weights published under a research license.

Test-time compute is the computation a model spends while answering a prompt, beyond what training already fixed into its weights. A 2024 study by Snell et al. showed how to spend it adaptively, and how it can stand in for a bigger model.

xAI's Grok-2 and Grok-2 mini, released in August 2024 as a beta for X subscribers, added image input, posted competitive benchmark scores and later had their weights published.

LMSYS (Large Model Systems) is a non-profit that started as a multi-university collaboration in 2023 and became an incorporated organisation in September 2024, building open models, serving systems and evaluation platforms such as Chatbot Arena, SGLang, FastChat and Vicuna.

Qwen2.5 is the Qwen family's September 2024 generation. Its launch included 13 open-weight models, with up to 18 trillion pretraining tokens, 128K context and most models under Apache 2.0.

Llama 3.2, announced on 25 September 2024, added vision models at 11 billion and 90 billion parameters and small text-only models at 1 billion and 3 billion for edge and mobile devices.

A data-center GPU announced by AMD on 10 October 2024, built on third-generation CDNA with 256 GB of HBM3E memory at 6 TB/s and aimed at models with a trillion parameters.

Yi-Lightning is 01.AI's flagship model from October 2024, built on an enhanced Mixture-of-Experts design. It ranked sixth overall on the LMSYS Chatbot Arena at launch and first among Chinese models.

The upgraded Claude 3.5 Sonnet, announced on 22 October 2024, raised coding and agentic scores and introduced computer use in public beta, the first such offer from a frontier AI model.

Hunyuan-Large is Tencent's open-weight mixture-of-experts model, with 389 billion total parameters and 52 billion activated, released in November 2024 under a Tencent license.

Lovable turns a plain-language description into a full-stack web application, aimed at people who cannot code. Launched in November 2024, it grew quickly in users and raised several large funding rounds at rising valuations.

Amazon Nova is Amazon's family of foundation models, introduced on 3 December 2024 and offered exclusively in Amazon Bedrock. It spans text understanding models from Micro to Pro, plus Canvas for images and Reel for video, and was succeeded by Nova 2 in December 2025.

AWS Trainium2 is Amazon Web Services' second-generation machine-learning chip, sold through 16-chip EC2 Trn2 instances and 64-chip Trn2 UltraServers, and pitched by AWS as offering better price performance than comparable GPU instances.

Gemini 2.0 Flash is the first model of Google's Gemini 2.0 generation, released as an experimental model in December 2024 with multimodal output and native tool calling, and made generally available in February 2025.

Phi-4 is Microsoft's 14-billion-parameter dense decoder-only language model, built with a data-quality-first recipe heavy on synthetic data and released with full weights under the MIT licence in January 2025.

Arena Intelligence is the company formed in 2025 by the team behind Chatbot Arena; it runs the LMArena model-comparison platform and states that it has 250 million dollars in venture funding.

MiniMax-01 is an open-weight model series released in January 2025. It includes the language model MiniMax-Text-01 and the vision-language model MiniMax-VL-01, and uses a hybrid attention design that supports very long context.

The DeepSeek app is the company's official free chatbot, offered on the web and as mobile apps since 15 January 2025. It topped the U.S. App Store in late January 2025.

iFlytek Spark X1 is a Chinese deep-reasoning model released in January 2025 and trained entirely on domestic computing hardware, with later upgrades benchmarked against OpenAI's reasoning models.

Kimi k1.5 is a multimodal reasoning model released by Moonshot AI on 20 January 2025, which the company said matched OpenAI's o1 in mathematics, coding and multimodal reasoning and which was trained with reinforcement learning rather than tree search or process reward models.

Doubao-1.5-pro, released by ByteDance on 22 January 2025, is a closed multimodal model built on a sparse mixture-of-experts design. It was offered through Volcano Engine at low API prices.

Qwen2.5-Max is Alibaba's large mixture-of-experts model, announced in January 2025. It was offered through Qwen Chat and the Alibaba Cloud API, and its weights were not released.

A paper showing that a 32B model fine-tuned on 1,000 carefully chosen questions and run with a simple "budget forcing" trick can beat o1-preview on competition maths by up to 27%, with model, data and code released.

OpenAI's small reasoning model released on 31 January 2025, offered to all ChatGPT users including the free tier, with three reasoning-effort levels and developer features such as function calling and Structured Outputs; succeeded by o4-mini in April 2025.

Thinking Machines Lab is a San Francisco AI company founded in February 2025 by former OpenAI CTO Mira Murati. It built Tinker, an API for fine-tuning open-weight models, and released the open-weight Inkling model.

Grok 3 and the cheaper Grok 3 mini are xAI's reasoning models from February 2025. Grok 3 added a Think mode, a 1-million-token context window and the DeepSearch research agent.

Claude 3.7 Sonnet, released on 24 February 2025, was described by Anthropic as the first hybrid reasoning model on the market, able to answer quickly or think step by step at length. It launched with Claude Code.

Alexa+ is Amazon's next-generation Alexa assistant, rebuilt on generative AI and introduced in February 2025. It costs $19.99 a month, is free for Prime members and can book services, order groceries and control smart-home devices.

Hunyuan Turbo S is a fast-thinking model released by Tencent on 27 February 2025, built on a hybrid Mamba-Transformer design with 560 billion total parameters. Its weights were not released.

GPT-4.5 is a large language model OpenAI released on 27 February 2025 under the code name "Orion". It was its largest chat model at the time, priced far above GPT-4o, and was retired from general availability within months.

Ling is a family of open-weight Mixture-of-Experts language models from Ant Group's InclusionAI, released with a technical report arguing that a 300B-scale MoE can be trained without premium GPUs.

A 32-billion-parameter open-weight reasoning model released by the Qwen team on 6 March 2025, trained mainly through scaled reinforcement learning and reported to perform comparably to the much larger DeepSeek-R1.

OpenAI's open-source toolkit for building multi-agent applications, released in March 2025 with configurable agents, handoffs, guardrails and tracing; it succeeds the experimental Swarm SDK and works with OpenAI and third-party models.

Gemma 3 is Google's third generation of open Gemma models, released on 12 March 2025 in 1B to 27B sizes, with a 128,000-token context window, image input on the larger variants, and support for more than 140 languages at the pretrained level.

Command A is Cohere's 111-billion-parameter model, released in March 2025 with a 256k-token context window. Cohere said it runs on two GPUs and delivers 150% higher throughput than Command R+ 08-2024.

ERNIE 4.5 is Baidu's native multimodal foundation model, released on 16 March 2025 and handling text, images, audio and video. In June 2025 Baidu open-sourced ten models from the family under the Apache License 2.0.

ERNIE X1 is Baidu's deep-thinking reasoning model released on 16 March 2025 alongside ERNIE 4.5, notable for tool use, multimodal input and low pricing, and later followed by X1 Turbo and X1.1.

Hunyuan-T1 is Tencent's deep-reasoning model, officially released on 21 March 2025 and built on the Hunyuan-TurboS base. Tencent did not release its weights.

Llama 4 Maverick is a mixture-of-experts model Meta released alongside Scout on 5 April 2025, with 400 billion total parameters and a 1-million-token context. Its ranking on LMArena was later questioned.

Llama 4 Scout is a mixture-of-experts model Meta released on 5 April 2025. It has 109 billion total parameters, 17 billion of them active for each token, and a context window of 10 million tokens.

Seed-Thinking-v1.5 is a reasoning model from ByteDance Seed: a 20B-activated, 200B-total mixture-of-experts model trained with reinforcement learning to think before answering, published with a technical report and benchmark tables but no model weights.

SenseTime's SenseNova V6 is a native mixture-of-experts multimodal foundation model with over 600 billion parameters, released on 10 April 2025 in general, reasoning and real-time interactive variants.

OpenAI's April 2025 family of three API models — GPT-4.1, GPT-4.1 mini and GPT-4.1 nano — offering a context window of over a million tokens, low prices per token and stronger coding scores. It was succeeded by GPT-5.

A small reasoning model released by OpenAI in April 2025 alongside o3, built for fast, low-cost reasoning; with a Python interpreter it reached 99.5% pass@1 on AIME 2025, and it was later succeeded by GPT-5 mini.

A model in Google's Gemini 2.5 family, released in preview in April 2025 and generally available that June. Google called it its first fully hybrid reasoning model, letting developers turn thinking on or off with a token budget.

MiMo-7B is a 7-billion-parameter reasoning model from Xiaomi's LLM-Core team, pre-trained on 25 trillion tokens and tuned with reinforcement learning on verifiable math and code problems, reported to surpass OpenAI o1-mini on several reasoning benchmarks.

Falcon-H1 is a family of open-weight language models that runs Transformer attention and state space layers in parallel, released in sizes from 0.5B to 34B parameters under the Falcon-LLM Licence.

Claude Opus 4 and Claude Sonnet 4, announced on 22 May 2025, were hybrid models with extended thinking. Anthropic called Opus 4 the best coding model in the world, and Claude Code became generally available the same day.

An updated release of DeepSeek's open-weight reasoning model, published on 28 May 2025 under the MIT License, with markedly higher scores on competition math, science and coding benchmarks, plus system-prompt support and stronger function calling.

MiniMax-M1 is an open reasoning model released by MiniMax in June 2025. It has 456B total parameters, a native 1-million-token context and an Apache 2.0 license, and it comes in 40K and 80K thinking-budget versions.

ByteDance Seed's general-purpose model series, announced on 11 June 2025 as Doubao Large Model 1.6, combining multimodal input, a 256K context window and a mode that decides per prompt whether to reason before answering.

AMD Instinct MI350X is a data-center GPU in AMD's MI350 series, built on the CDNA 4 architecture with up to 288 GB of HBM3E memory, up to 8 TB/s of bandwidth and a 1,000 W board power.

GLM-4.5 is Zhipu's open-weight language model, released in July 2025 with 355 billion total parameters and a lighter sibling, GLM-4.5-Air. Both are hybrid reasoning models released under the MIT License.

Grok 4 and Grok 4 Heavy are xAI's models from July 2025. Grok 4 has a 256,000-token context and was trained with reinforcement learning at pretraining scale, while Heavy runs parallel test-time compute.

EXAONE 4.0 is LG's family of large language models, released as a 32B model and a 1.2B on-device model, merging chat usability with a reasoning mode and adding tool use and Spanish support.

Qwen3-2507 is a staged 2025 refresh of the open-weight Qwen3 models that split thinking and non-thinking into separate checkpoints, raised native context to 262,144 tokens, and scored far above the original non-thinking mode on published benchmarks.

A 14.74-billion-parameter Korean-English reasoning model released by NAVER in July 2025, built by compressing a larger model and training it with a four-stage reasoning recipe at a fraction of the cost of comparable models.

An enhanced reasoning mode for Gemini 2.5 Pro that weighs multiple hypotheses in parallel before answering, announced experimentally in 2025 and known for reaching gold-medal standard at the International Mathematical Olympiad.

gpt-oss is a family of open-weight reasoning language models released by OpenAI on 5 August 2025, in 120-billion and 20-billion parameter sizes, under the permissive Apache 2.0 license.

ByteDance Seed's open-weight 36-billion-parameter model family from August 2025, trained on 12 trillion tokens with a 512K-token context window and a user-settable thinking budget.

DeepSeek-V3.1, released on 21 August 2025 under the MIT License, was DeepSeek's first hybrid model, supporting both a thinking mode and a non-thinking mode in one set of weights. It has 671 billion total parameters, with 37 billion activated per token.

Qwen3-Max is Alibaba's proprietary trillion-parameter language model, released in September 2025 and served through Alibaba's API. It is the Max-tier flagship that followed Qwen2.5-Max.

LongCat-Flash is a 560-billion-parameter Mixture-of-Experts language model from Meituan's LongCat team, which activates roughly 27 billion parameters per token and was released with open weights under the MIT license.

Claude Sonnet 4.5, released on 29 September 2025, was a drop-in replacement for Sonnet 4 at the same price. Anthropic reported that it could stay focused on complex tasks for more than 30 hours.

GPT-5.1 is an update to GPT-5 released by OpenAI on 12 November 2025. It added a warmer default personality with selectable styles, and its API version has been scheduled for shutdown in 2027.

ERNIE 5.0 is Baidu's natively omni-modal foundation model, unveiled in November 2025 and described as having 2.4 trillion parameters. It is proprietary and was made available to enterprises and consumers in January 2026.

Olmo 3 is the third generation of the Allen Institute for AI's fully open language models, announced on 20 November 2025 in 7B and 32B dense Transformer sizes, with a 65,536-token context and all datasets openly available.

Claude Opus 4.5, released on 24 November 2025, cost a third as much as Opus 4.1 per token and introduced an effort parameter that lets developers trade capability for time and cost. Claude Opus 4.6 succeeded it.

DeepSeek-V3.2, released on 1 December 2025, succeeded the experimental V3.2-Exp and integrated thinking into tool use. Its companion V3.2-Speciale, which maximizes reasoning, reached gold-level results at several 2025 competitions.

Mistral AI's flagship model of the Mistral 3 family, released in December 2025: a sparse mixture-of-experts model with 675B total parameters, open weights under Apache 2.0, and image understanding.

Gemini 3 Deep Think is Google's enhanced reasoning mode for Gemini 3, announced with Gemini 3 Pro on 18 November 2025 and released to Google AI Ultra subscribers on 3 December 2025, with an upgraded version following in February 2026.

Tencent's flagship model released in December 2025, offered in Think and Instruct versions, built on a 406-billion-parameter mixture-of-experts architecture with a 256K-token context window.

Solar Open 100B is Upstage's open-weight mixture-of-experts language model for English and Korean, released in December 2025 with 102.6 billion total parameters, 12 billion active per token, and a custom licence.

GPT-5.2 is a model family OpenAI released on 11 December 2025, with Instant, Thinking and Pro versions and a coding variant. It was aimed at professional work such as spreadsheets and presentations.

Gemini 3 Flash is a fast, lower-cost model in Google's Gemini 3 family, released on 17 December 2025. It scored close to Gemini 3 Pro on several benchmarks while using fewer tokens and costing far less per token.

The AMD Instinct MI400 is AMD's next-generation data-center GPU for AI, expected in 2026 with 432 GB of HBM4 memory, and designed to be deployed inside AMD's Helios rack-scale system.

Open Responses is an open specification and ecosystem, based on the OpenAI Responses API, that gives applications one shared schema for calling models, streaming results and composing agentic workflows across providers.

Microsoft's Maia 200 is an inference accelerator introduced on 26 January 2026, built on TSMC's 3-nanometer process with 216 GB of HBM3e memory and more than 10 petaflops at FP4, aimed at serving models such as OpenAI's GPT-5.2.

Step 3.5 Flash is an open-weights Mixture-of-Experts language model from StepFun, released in February 2026 with 196 billion total parameters, a 256K-token context and an Apache 2.0 license.

Released on 5 February 2026, Claude Opus 4.6 was an upgrade to Anthropic's smartest model. It was the first Opus-class model with a 1M-token context window, in beta at launch, and it introduced agent teams in Claude Code.

GPT-5.3-Codex is a coding model OpenAI released on 5 February 2026, which the company called its first model that was instrumental in creating itself and pitched for repository search, terminal commands and debugging as well as writing code.

GLM-5 is Zhipu's open model released on 12 February 2026 under the MIT License, with 744 billion parameters and 40 billion active. It was built for agentic engineering.

Seed2.0 is ByteDance's February 2026 Doubao model series, offered in three agent-oriented sizes with low token pricing and reported benchmark results.

Qwen3.5 is a natively multimodal model family from Alibaba released in February 2026. Its open flagship, Qwen3.5-397B-A17B, has 397B total parameters and 17B active per token, under Apache 2.0.

Released on 17 February 2026 with a 1M-token context window in beta, Claude Sonnet 4.6 kept the price of Sonnet 4.5 and became the default model for Free and Pro users on claude.ai.

Gemini 3.1 Pro is Google's preview upgrade to Gemini 3 Pro, released on 19 February 2026, scoring a verified 77.1% on ARC-AGI-2 and shipped through the Gemini API, the Gemini app and NotebookLM.

GPT-5.4 is a model OpenAI released on 5 March 2026 in Thinking and Pro versions. It added built-in computer use and improved deep research, and it was the model that succeeded GPT-5.2.

Gemma 4 is Google's fourth generation of open-weight models, released in April 2026 and built on the same research and technology as Gemini 3, ranging from phone-sized E2B and E4B to a 26B mixture-of-experts and a 31B dense model.

A generation of Alibaba's Qwen models released in April 2026, spanning proprietary Plus and Max versions aimed at agentic coding and open-weight models such as Qwen3.6-35B-A3B, a 35B-parameter mixture-of-experts with 3B active parameters.

Muse Spark is the first model from Meta Superintelligence Labs, released on 8 April 2026 as a closed model that powers the Meta AI app and meta.ai. An open-weight sibling followed later.

EXAONE 4.5 is LG AI Research's first open-weight vision language model: a 33-billion-parameter model with a 1.2-billion-parameter vision encoder, trained on documents and text, with a 256K-token context window, released in April 2026.

Released on 16 April 2026 at the same price as Opus 4.6, Claude Opus 4.7 added an xhigh effort level and accepted much larger images. It was the first model to test automatic blocking of high-risk cybersecurity requests.

The Qwen3.5-Omni Technical Report documents Alibaba's Qwen3.5-Omni, a hundreds-of-billions-parameter model that understands long audio and video and generates speech across 10 languages, reporting state-of-the-art results on 215 audio and audio-visual benchmarks.

GPT-5.5 is an OpenAI model released on 23 April 2026 as GPT-5.5 Thinking and Pro, with an API version the next day. It succeeded GPT-5.4 and had a cybersecurity variant in limited preview.

Laguna XS.2 is Poolside's first open-weight coding model: 33 billion total parameters with 3 billion active per token, released under Apache 2.0 and trained on 30 trillion tokens.

Laguna M.1 is Poolside's 225-billion-parameter mixture-of-experts model for long-horizon agentic coding, activating 23 billion parameters per token. Its weights reached Hugging Face seven weeks after launch.

Meta's safety and preparedness report for Muse Spark, the model behind Meta AI, sets out the catastrophic-risk evaluations behind its launch and argues that, with mitigations applied, residual risk is acceptable.

Gemini 3.5 Flash is Google's first Gemini 3.5 model, announced on 19 May 2026 and built for agentic and coding work; it became the default model in the Gemini app and in AI Mode in Search.

Command A+ is Cohere's open-source mixture-of-experts language model, released on 20 May 2026 under the Apache 2.0 licence with 218 billion total and 25 billion active parameters, aimed at sovereign and critical-infrastructure deployments.

MiniCPM5 is a series of small, dense, open-weight language models from the MiniCPM team, beginning with a 1B model released in May 2026 and followed by a 2B model in September 2026, all aimed at on-device and other resource-constrained deployment.

A mixture-of-experts language model released by MiniMax in June 2026, with about 428B total parameters, a 1M-token context window and a new sparse attention scheme built for long-context and agent-style work. [^1]

MAI-Thinking-1 is Microsoft AI's flagship reasoning model, launched on 2 June 2026 at Microsoft Build. Microsoft said it was preferred to Anthropic's Claude Sonnet 4.6 in blind human evaluations, and that developers could fine-tune its weights for the first time.

Zhipu's open-weight language model released in June 2026 under the MIT License, with a 1M-token context, a shared sparse-attention indexer, and higher reasoning and coding scores than GLM-5.1.

ByteDance Seed's Seed2.1, released in June 2026 in Pro and Turbo versions, is a Doubao-family model aimed at real productivity work, with reported gains in agent ability, coding, knowledge, reasoning and multimodal understanding.

Claude Sonnet 5, released on 30 June 2026, is Anthropic's most agentic Sonnet model. It can make plans, use browsers and terminals, and run autonomously, at an introductory price of $2 and $10 per million tokens.

Tencent Hy3 is an open-weight reasoning model released on 6 July 2026 under Apache 2.0, with 295 billion total parameters, 21 billion active and a 256K context.

OpenAI's three-model family released on 9 July 2026 — Luna, Terra and Sol, ordered from least to most capable — with Sol pitched as the company's strongest coding and agent model.

Inkling is the first model released by Thinking Machines Lab: a 975B-parameter mixture-of-experts model with open weights, native text, image and audio input, and a context window of up to 1M tokens.

Claude Opus 5, released on 24 July 2026 at $5 and $25 per million tokens, is Anthropic's most aligned model to date and its most capable generally available model for scientific research.

Qwen3.8 is Alibaba's 2026 flagship line. Its open 2.4-trillion-parameter model, Qwen3.8-2.4T-A95B, released in August 2026, has 95B active parameters and a custom license.

Tencent Hy4 preview is an open-weight mixture-of-experts model released on 28 August 2026 with 770 billion total parameters, 49 billion active and a context of over one million tokens.

Step 5 Preview is a preview release of StepFun's flagship foundation model: a sparse mixture-of-experts system with 600 billion total parameters, a 1-million-token context window and published weights, released in September 2026.

Claude Fable 5.1, released on 1 September 2026, keeps the $10 and $50 per million token price of Fable 5, with cache reads 75% cheaper. Its sibling Claude Mythos 5.1 is limited to trusted-access programmes.

Gemini 3.8 Flash is Google's fast, low-cost model released on 2 September 2026, alongside a cybersecurity variant, Gemini 3.8 Flash Cyber, with introductory pricing that doubles at the start of 2027.

DeepSeek-V4.1-Flash, released on 10 September 2026 under the API name deepseek-flash, is a 552-billion-parameter multimodal mixture-of-experts model with a causal encoder-decoder design and a 1M-token context.

Claude Opus 5.5, released on 22 September 2026, is the first model in Anthropic's Claude 5.5 family. Anthropic said it performs at the level of Fable 5.1 on most work, at about 40% lower cost than Opus 5.

Claude Sonnet 5.5, released on 28 September 2026, is the second model in the Claude 5.5 family and Anthropic's fastest Sonnet to date. It costs $2 and $10 per million tokens, with cache reads halved in October 2026.

OpenAI's late-2026 model family: Sol, released at DevDay on 29 September 2026, and the faster, cheaper Terra on 26 October 2026. A planned Astra version was withheld over safety concerns.

Kimi K3.1 Pro is Moonshot AI's October 2026 open-weights model, the first in the K3.1 series, built for financial research, 3D modeling and coding. It keeps K3's mixture-of-experts scale but swaps in sparse attention layers.

Kolibri-1 is Aleph Alpha's open-weight German–English mixture-of-experts model: 78.1 billion parameters with 3.46 billion active per token, a 1,048,576-token context window, and weights that fit on a single H200 or B200.

Mistral Large 4 is Mistral AI's flagship open-weight model, with about 1 trillion total parameters, a 1M-token context window and native multimodality. It was previewed on 6 October 2026, and an improved version's weights followed on 29 October 2026.

Claude Haiku 5.5, released on 7 October 2026, is the third model in the Claude 5.5 family. Anthropic called it the cheapest, fastest and most capable small model it had released, at about 75% less cost to run on average.

Muse Spark Open is Meta's open-weight large language model, announced on 15 October 2026. With 816 billion total parameters, it is built for coding and agentic work, and its weights can be downloaded, unlike the closed Muse Spark that started the family.

Jais-3-MoVA-36B-A4B is an Arabic-centric mixture-of-experts model from the Institute of Foundation Models: 36 billion parameters, 4 billion active per token, a 512,000-token context and a mixture-of-values attention scheme.

Jais-3-7B is a dense 7B Arabic-and-English model from the Institute of Foundation Models, with a 512,000-token context window, a single-GPU footprint of about 18 GB, and reasoning returned in a separate field.

LFM3-Omni-2B is Liquid AI's first omni-modal model in the LFM3 family: one 2.17-billion-parameter model that reads text, images, video and audio and answers in text or speech.

iFlytek's Spark X2.6 is a device-cloud model family unveiled in October 2026, pairing a 293-billion-parameter mixture-of-experts cloud model with smaller on-device models and a router that decides where each task runs.

Gemini 4 Carbon is a Google DeepMind frontier model announced on 28 October 2026, two weeks after Gemini 4 Argon, with Google claiming state-of-the-art cybersecurity and long-horizon capabilities.

OpenAI's cybersecurity-focused model, released on 5 November 2026 and limited to the top tier of its Daybreak Red programme; OpenAI says it surfaced 948 potential Linux kernel issues and 81 novel Firefox vulnerabilities.

Aidan Gomez is a Canadian computer scientist, co-founder and CEO of Cohere, and one of the eight authors of the 2017 paper that introduced the Transformer architecture.

The Technology Innovation Institute is a research centre of Abu Dhabi's Advanced Technology Research Council that works across ten technology fields and develops the Falcon family of language models, releasing model weights, technical reports and its own licence.

Red teaming is the practice of actively probing a model for harmful or unwanted behaviour before deployment. A 2022 method by Perez and colleagues automated it by using one language model to generate test cases for another.

Fidji Simo is a technology executive who led Instacart and became OpenAI's chief executive of Applications in 2025; in July 2026 she stepped down for health reasons and moved into a part-time advisory role.

Stanford computer science professor who directs the Center for Research on Foundation Models and works on making language models more open and more rigorously benchmarked.

A jailbreak is an adversarial prompt that gets a safety-trained model to produce behaviour its training was meant to prevent. A 2023 analysis by Wei et al. traced these failures to competing objectives and mismatched generalization.

Christopher Manning is a Stanford professor of linguistics and computer science, former SAIL director, and former ACL president, recognized with fellowships and academy memberships.

Sarvam AI is an Indian company building a sovereign AI stack; its Sarvam-M, a post-trained 24-billion-parameter Mistral Small released under Apache-2.0, reports large gains on Indian-language benchmarks.

Ant Group is the company whose InclusionAI team released the open-weight Ling (Bailing) family of mixture-of-experts language models in 2025, including Ling-Lite and Ling-Plus, trained without premium GPUs and published under the MIT licence.

Noam Brown is an AI researcher at OpenAI, listed on his homepage as working on reasoning, reinforcement learning, self-play and multi-agent AI. He is known for the poker systems Libratus and Pluribus and for work on CICERO and reasoning models such as OpenAI o1.

Snowflake is a cloud data platform company that also builds its own open language models; its Arctic model pairs a dense transformer with a sparse mixture of experts and shipped under Apache 2.0.

Transformers is an open-source Python library that offers pretrained Transformer models through one unified API. Its paper was posted to arXiv on 9 October 2019, and the first v5 release candidate came in December 2025.

Alec Radford is a machine learning researcher who worked at OpenAI from 2016 to 2024. He was first author of the GPT-2, CLIP and Whisper papers, and lead author of the 2018 paper that introduced the GPT approach.

Noam Shazeer is an AI researcher who co-wrote the Transformer paper and was first author of the paper on sparsely gated mixture-of-experts layers. He co-founded Character.AI, returned to Google in 2024, and announced a move to OpenAI in 2026.

Sierra is a company co-founded by Bret Taylor and Clay Bavor that helps businesses build customer experiences with AI agents, positioning the technology as augmentation of human staff rather than replacement.

A context window is the full span of text a model can refer to while producing a response, measured in tokens and including the response itself. Its size, and how evenly a model uses it, shape what the model can do.

Long context is a model's capacity to take in and work over very large inputs — documents, images, recordings — in a single request. Systems now reach millions of tokens, though finding information inside that window is a separate skill.

Meituan's LongCat team released LongCat-Flash, a 560-billion-parameter mixture-of-experts language model that activates roughly 27 billion parameters per token on average and ships with open weights under the MIT licence.

A KV cache keeps the key and value vectors a model has already computed for earlier tokens, so each new token reuses them instead of recomputing. Its memory footprint is what makes high-throughput serving hard to scale.

Prompt engineering is the practice of designing the text input that steers a generative AI system toward a desired output. Shared vocabulary and techniques such as chain-of-thought prompting make results easier to describe, compare and reproduce.

Ashish Vaswani is a computer scientist best known as the first-listed author of “Attention Is All You Need,” the 2017 paper that introduced the Transformer architecture. He also co-founded the AI companies Adept and Essential AI.

Aleph Alpha is the company behind Kolibri, an open-weight model released on 3 October 2026 and built around data sovereignty and German-language specialisation rather than raw scale.

LG AI Research is the organisation behind EXAONE, a family of large language models it documents in technical reports and releases under its own EXAONE licence.

Jan Leike is an AI alignment researcher who joined Anthropic in May 2024 after leaving OpenAI, where he had co-led the superalignment project with Ilya Sutskever.

Jason Wei is an AI researcher and first author of papers on chain-of-thought prompting and emergent abilities in large language models; he has worked at Google Brain, OpenAI and Meta Superintelligence Labs.

Liquid AI is an MIT CSAIL spin-out that calls itself an efficiency-first foundation model company, arguing that models should run on local devices; its LFM3-Omni-2B is an omni-modal model published in October 2026.

Poolside is an AI company building models for software work; in April 2026 it released the Laguna M.1 and Laguna XS.2 models and pool, the agent runtime behind its reinforcement learning.

Epoch AI's Benchmarking Hub collects benchmark results for leading AI models, including Epoch-run evaluations, and combines many scores into the Epoch Capabilities Index, a single general capability scale.

Synthetic data is training material written by a model rather than collected from people or the web. In the phi-1 experiments a small code model trained partly on model-generated "textbooks" reached competitive benchmark scores, suggesting data quality itself shapes capability.

Tool calling lets a language model hand a sub-task, such as arithmetic or search, to an external program and use the result in its answer. Toolformer showed in 2023 that a model could learn when and how to call tools on its own.

Planning in AI agents is the process of breaking a goal into intermediate steps, choosing actions, and revising the plan as results arrive, rather than producing an answer in one pass.[^1][^2]

Jared Kaplan is a physicist turned AI researcher who co-founded Anthropic and serves as its Chief Science Officer. He was first author of the 2020 paper “Scaling Laws for Neural Language Models.”

Reka AI is an AI research lab building models and infrastructure for the physical AI era. Its releases include reka-flash-3, a 21-billion-parameter reasoning model, and Reka Edge, a 7-billion-parameter vision-language model designed to run on devices at the edge.

Benchmark contamination means the test items a model is evaluated on, or close variants of them, were present in its pretraining data, so scores partly measure memory of seen text rather than the ability the benchmark is meant to probe.

A French-language public research university in Montreal, founded in 1878 and autonomous since 1919, known today as a centre of deep-learning research and the home institution of Yoshua Bengio and Mila.

A public research university in Beijing's Haidian district, founded in 1911 as Tsinghua College. It is now a major centre for computer science and AI research, with which Turing laureate Andrew Yao, ResNet author Kaiming He and spin-out Zhipu AI are associated.

Yoav Shoham is an Israeli-American computer scientist, professor emeritus at Stanford and co-founder and chairman of AI21 Labs. He initiated the AI Index and has worked on AI, game theory and multi-agent systems.

Liu Qingfeng is the president and chairman of iFlytek, the company he founded in 1999 while a Ph.D. student at USTC. He unveiled the Spark large model in May 2023.

Chinese computer scientist and entrepreneur who built Sogou Search and the Sogou input method, led Sogou as CEO through its New York listing, and later founded the AI company Baichuan.

Spanish computer scientist (born 1983) who co-invented sequence-to-sequence learning, led DeepMind's AlphaStar research group, and became a technical lead on Gemini before leaving to start Discovery Loop.

A 1993 journal article by Mitchell P. Marcus, Beatrice Santorini and Mary Ann Marcinkiewicz describing the Penn Treebank: over one million words of Wall Street Journal text annotated with syntactic bracketing, built as a resource for work on English sentence structure.

Canadian machine-learning researcher who was one of Geoffrey Hinton's earliest hires at Google Brain's Toronto office, co-founded the natural language processing company Cohere in Toronto in 2019, and is lead vocalist of the indie rock band Good Kid.

American entrepreneur born in 1997 who co-founded Scale AI, a data-labelling and model-evaluation company, and became Meta's chief AI officer and leader of Meta Superintelligence Labs in 2025.

XiaoIce was a Microsoft chatbot released in May 2014 that users added to WeChat group chats for reminders, lookups, weather and jokes. It drew large crowds, then was banned by WeChat three days later for breaking platform rules.

Cerebras Systems is an American AI chip company founded in 2015 that builds wafer-scale processors — single chips made from a whole silicon wafer — and the computers around them, from the WSE and CS-1 to the WSE-3 and CS-3.

HellaSwag is a 2019 benchmark in which a machine must pick the most plausible continuation of a short event description. Its wrong answers are generated and filtered to fool models rather than people.

ERNIE 2.0 is Baidu's pre-trained language model framework, published in 2019, which builds pre-training tasks incrementally through continual multi-task learning and reported results above BERT and XLNet on English and Chinese benchmarks.

A 2019 Google paper that frames every language task as text in, text out, using one encoder-decoder Transformer and one training setup. It also introduced the C4 dataset, and the released T5 models became common research starting points.

Writer is an enterprise AI company founded in 2020 by May Habib and Waseem AlShikh, selling brand-aware agents and governance tools to large companies, and publishing the Palmyra-mini family of small open models.

Meena is a 2.6-billion-parameter open-domain chatbot from Google, trained end-to-end on filtered social media conversations. Its 2020 paper introduced the SSA human-evaluation metric and linked next-token perplexity to human-judged chat quality. The work was later renamed LaMDA.

GPT-Neo is a family of open text-generation models released by EleutherAI in March 2021 at 125M, 1.3B and 2.7B parameters, built to replicate the GPT-3 architecture and distributed with freely downloadable weights under an MIT license.

Langboat Technology is a Chinese AI company founded in June 2021 and incubated at Sinovation Ventures; it develops the in-house Mengzi model family and sells an enterprise agent platform and applications built on it.

Jurassic-1 is the first family of large language models from AI21 Labs, announced with the AI21 Studio developer platform in August 2021. Its largest model, Jumbo, has 178 billion parameters.

Megatron-Turing NLG 530B is a 530-billion-parameter language model announced by NVIDIA and Microsoft on 11 October 2021, then the largest monolithic transformer language model, trained on NVIDIA's Selene supercomputer with Megatron-LM and DeepSpeed combined.

Character.AI is a chatbot company founded in 2021 by former Google engineers Noam Shazeer and Daniel De Freitas, known for user-created AI characters; it filed for Chapter 11 bankruptcy in October 2026.

Gopher is DeepMind's 280-billion-parameter language model, announced in December 2021 alongside a study of what scale does to language models and a companion taxonomy of their risks.

ERNIE 3.0 Titan is a Chinese pre-trained language model with over 260 billion parameters, described by its authors as the largest dense Chinese model of its time and reported to beat state-of-the-art systems on 68 NLP datasets.

A Chinese AI company founded in 2022 under Tsinghua associate professor Liu Zhiyuan, known for the MiniCPM on-device model series, multimodal and speech models, and the ForgeTrain training framework.

AlphaCode is a DeepMind code-generation system for competitive programming, announced in February 2022. On Codeforces contests held after its training data, it reached an estimated rank around the median participant. A Gemini-based AlphaCode 2 followed in December 2023.

Inflection AI is an American AI company founded in 2022 by Reid Hoffman, Mustafa Suleyman and Karén Simonyan. It built the chatbot Pi and, in 2024, moved most of its staff and technology to Microsoft.

Flamingo is a DeepMind visual language model announced in April 2022. It takes prompts that mix images, videos and text and replies in language, learning new tasks from only a few examples.

OPT-175B is a 175-billion-parameter language model from Meta AI, announced in May 2022. Meta said it was the first model of its size released with both pretrained weights and the code to train and use them.

Speculative decoding speeds up text generation from large autoregressive models: a smaller model proposes several tokens and the large model checks them in parallel, producing identical output faster.

Galactica is a large language model for science, unveiled by Meta AI and Papers with Code in November 2022. Its public demo was taken down about three days later, after users found it produced fake papers.

A pair of medical question-answering language models described in arXiv papers of 26 December 2022 and 16 May 2023; the first introduced the MultiMedQA benchmark, and Med-PaLM 2 reached up to 86.5% on MedQA.

The consumer chat product of Zhipu AI, launched as ChatGLM in 2023 and later renamed GLM. It is the web entry point where new Zhipu models such as GLM-5.1 reach ordinary users.

Cohere's first flagship instruction-following generative model, released in 2023 and distributed through Cohere's API and cloud marketplaces such as Amazon SageMaker JumpStart and Oracle Cloud Infrastructure, later superseded by Command R and deprecated in 2025.

Text watermarking embeds a hidden statistical signal in generated text so that a short span of tokens can later be identified as machine output, while the text still reads normally to people.

Jurassic-2 is a family of large language models released by AI21 Labs in March 2023 as the successor to Jurassic-1, offered in three sizes with instruction-tuned variants, multilingual support and task-specific APIs; it was succeeded by Jamba in 2024.

A lighter, cheaper and faster version of Anthropic's Claude model, released in 2023 for business use through the API and retired in November 2024.

RedPajama is an open-source project that began by reproducing the LLaMA training dataset, releasing a 1.2-trillion-token corpus as a clean-room, fully open implementation.

Pythia is a suite of 16 open language models from EleutherAI, ranging from 70 million to 12 billion parameters, all trained on the same data in the same order, with 154 public checkpoints from each training run.

SenseChat is SenseTime's conversational AI product, launched on 10 April 2023 as part of the SenseNova model family with more than 100 billion parameters, first offered to government and corporate clients and later expanded through the V4 and SenseNova V6 releases.

Amazon Titan is a family of foundation models built by Amazon for Amazon Bedrock, covering text generation, embeddings and image generation. It was announced in 2023 and later succeeded by Amazon Nova.

Stable LM is Stability AI's family of open language models, first released in April 2023 as 3B- and 7B-parameter alpha versions under licenses that allow commercial use of the base models and research-only use of the instruction-tuned ones.

HuggingChat is a free chatbot launched by Hugging Face on 25 April 2023 as an open-source alternative to ChatGPT. It initially ran the volunteer Open Assistant model and let users switch among open models.

Pi is a chatbot launched by Inflection AI on 2 May 2023, named for "Personal Intelligence" and presented as a personal, conversational companion. Most of Inflection's staff, including two co-founders, moved to Microsoft in March 2024.

StarCoder is an open-access code language model of about 15.5 billion parameters, released on 4 May 2023 by the BigCode project led by Hugging Face and ServiceNow.

Vercel's AI SDK is a TypeScript toolkit that gives developers one interface for many AI model providers, framework-agnostic hooks for chat interfaces, and a uniform way to run agent harnesses.

Baichuan-7B is an open-source, 7-billion-parameter language model for Chinese and English released in June 2023, the first model from Baichuan, trained on about 1.2 trillion tokens and evaluated on Chinese and English benchmarks.

ERNIE 3.5 is a Baidu large language model announced in June 2023 as the upgrade behind the ERNIE Bot chatbot, adding a plugin architecture and, according to Baidu, outperforming ChatGPT on several benchmarked abilities.

Phi-1 is a 1.3-billion-parameter Transformer language model for Python from Microsoft Research, which showed that filtered and synthetic "textbook quality" training data can partly substitute for model and data scale.

An open-source 13-billion-parameter language model from Baichuan, released in Base and Chat versions in July 2023, trained on 1.4 trillion tokens and benchmarked against Meta's LLaMA-13B.

Skylark, or Yunque (云雀), is ByteDance's first large language model. Released in August 2023, it powered the Doubao chatbot and was relaunched in May 2024 as the Doubao model family.

SparkDesk V2.0 is the second version of iFlytek's large language model, launched on 15 August 2023 with upgraded coding and multimodal abilities, and released alongside the iFlyCode 1.0 programming assistant.

Phi-1.5 is a 1.3-billion-parameter language model from Microsoft Research, trained on curated synthetic "textbook" text rather than web crawls; its report claims natural-language performance comparable to models five times larger.

PLaMo-13B is a 13-billion-parameter, LLaMA-based language model from Preferred Networks, pre-trained on English and Japanese open data and released on Hugging Face in September 2023 under the Apache-2.0 licence.

iFlytek's Spark V3.0, released on 24 October 2023, is a large language model the company said beat GPT-3.5 in Chinese; it arrived with a domestic Huawei-built computing platform and twelve industry models.

DeepSeek's first model release: a family of open-source code models from 1.3B to 33B parameters, trained from scratch on 2 trillion tokens and reported to beat Codex and GPT-3.5 on several benchmarks.

DeepSeek LLM is DeepSeek's first general-purpose language model series, released on 29 November 2023 in 7B and 67B sizes, each with a base version and a chat version.

Phi-2 is a 2.7-billion-parameter base language model from Microsoft Research, announced on 12 December 2023. Microsoft reported that it matched or beat models up to 25 times larger on complex benchmarks, making it a reference point for small-model training.

DeepSeekMoE is a mixture-of-experts architecture from DeepSeek, released in January 2024, that splits experts finely and keeps a few shared experts always active so that each expert holds focused knowledge. Its 16B open-weights model matched larger dense models at a fraction of the computation.

Baichuan 3 is a closed language model with more than one hundred billion parameters, released by Baichuan on 29 January 2024. The company reported that it matched or beat GPT-4 on several Chinese benchmarks and on Chinese medical tasks.

Spark V3.5 is a large language model that iFlytek released on 30 January 2024. It was the first Spark model trained entirely on the domestic FlyingStar One computing platform, and iFlytek compared it with GPT-4 Turbo.

Dolma is an openly released English corpus of three trillion tokens, built from web text, papers, code, books, social media and Wikipedia, with a paper documenting both the dataset and the pipeline used to curate it.

Aya 101 is a 13-billion-parameter, instruction-tuned language model from Cohere For AI that follows instructions in 101 languages. It was released under Apache 2.0 in February 2024.

DeepSeekMath is a 7B-parameter open language model built for mathematical reasoning and released in 2024. It reached 51.7% on the competition-level MATH benchmark without external tools, and introduced GRPO, a reinforcement-learning algorithm that uses less memory than PPO.

Qwen1.5 is the Qwen family's open model release of February 2024, with base and chat models from 0.5B to 72B parameters, later extended to 110B, and a mixture-of-experts model.

SenseNova 4.0, announced on 8 February 2024, upgraded SenseTime's model family with 4K, 32K and 128K context versions, specialised models and a first multimodal tool-calling API.

StarCoder2 is a family of open code models in 3B, 7B and 15B sizes, released in February 2024 by BigCode (Hugging Face and ServiceNow) with NVIDIA and trained on The Stack v2, published together with its dataset and training code.

Cohere's 35-billion-parameter language model announced in March 2024, built for retrieval-augmented generation and tool use in production, with a 128k-token context window and weights released for research.

Step-1V is a roughly 100-billion-parameter multimodal model from StepFun, unveiled in March 2024, that reads text, data and charts in images and supports video understanding; it ranked first on OpenCompass at launch and was succeeded by Step-1.5V.

Step-1 is StepFun's first language model, a system of roughly one hundred billion parameters unveiled in Shanghai on 23 March 2024 alongside the multimodal Step-1V and a preview of Step-2.

Grok-1.5 is xAI's upgrade to Grok-1, with better reasoning and a 128,000-token context window. Its companion Grok-1.5V was xAI's first multimodal model. Neither model's weights were released publicly.

Mengzi3 is an open-source large language model series from Langboat, released in 8B and 13B sizes under Apache-2.0, built on the Llama architecture and trained further on trillions of multilingual tokens.

A sparse mixture-of-experts language model from Mistral AI, released in April 2024 under the Apache 2.0 licence with 141 billion total parameters, of which 39 billion are active per token.

abab 6.5 is a mixture-of-experts language model family released by MiniMax in April 2024, reported to support a 200,000-token context and to answer all 891 queries in a long-context retrieval test.

OpenELM is a family of small open language models from Apple, published in four sizes from 270M to 3B parameters with its training framework, logs, checkpoints and MLX conversion code.

Yi-1.5 is an open-source family of language models from 01.AI, released in 34B, 9B and 6B sizes in base and chat forms, built by continuing to pre-train the earlier Yi models on a large new corpus.

Gemini 1.5 Flash is a lightweight, fast model in Google's Gemini 1.5 family, announced on 14 May 2024 and trained by distillation from 1.5 Pro, with a 1-million-token context window.

Baichuan 4 is a closed large language model released by the Chinese company Baichuan on 22 May 2024, the company's first multimodal model and the engine behind its first AI assistant, Baixiaoying.

Aya 23 is a family of open-weight multilingual instruction-tuned models from Cohere For AI, released in 8-billion and 35-billion-parameter sizes in May 2024. It covers 23 languages and was reported to outperform Aya 101 and several contemporary open models.

Codestral is Mistral AI's family of code models, first released in May 2024 as an open-weight 22B-parameter model with a 32K context window, followed by faster and longer-context versions aimed at IDE code completion.

ERNIE 4.0 Turbo is a large language model Baidu released in June 2024 as a faster, cheaper upgrade to ERNIE 4.0, sitting between ERNIE 4.0 and ERNIE 4.5 in the company's model line.

Apple Foundation Models (AFM, 2024) are the first generation of the models behind Apple Intelligence: a roughly 3-billion-parameter on-device model and a larger server model that runs in Private Cloud Compute.

Nemotron-4 340B is a family of open models from NVIDIA, released on 14 June 2024 as base, instruct and reward models. It was built to generate synthetic data for training other language models, and it fits on a single eight-GPU DGX H100.

An open-source mixture-of-experts code model from DeepSeek, released in June 2024, that widened language coverage and context length over its predecessor and reported benchmark results above several closed models.

Google's second-generation Gemma open models, announced on 27 June 2024 in 9-billion and 27-billion-parameter sizes, using grouped-query attention, an 8,192-token context window and a license permitting commercial use.

SenseTime's July 2024 model family, unveiled at WAIC, adding a real-time multimodal model, lightweight and edge versions, and migration incentives aimed at OpenAI customers.

SmolLM is a family of small language models with 135M, 360M and 1.7B parameters, released openly by Hugging Face in July 2024 for on-device use and trained on a purpose-built corpus of generated textbooks, code and filtered web text.

Mistral NeMo is a 12-billion-parameter language model released by Mistral AI with NVIDIA in July 2024, notable for a 128K-token context window, a new multilingual tokenizer called Tekken, and an Apache 2.0 release.

Phi-3.5 is Microsoft's August 2024 update to its small Phi models, released in three versions — mini, vision and the line's first mixture-of-experts model — under the MIT licence.

EXAONE 3.0 7.8B is the first openly released model in LG AI Research's EXAONE family: an instruction-tuned model strong in Korean, published with gated weights on Hugging Face.

Jamba 1.5 is AI21 Labs' family of instruction-tuned language models, Mini and Large, which combine Transformer and Mamba layers, offer a 256,000-token effective context and ship with open weights.

OLMoE is a fully open sparse mixture-of-experts language model from Ai2: 7 billion total parameters, 1 billion active per token, pretrained on 5 trillion tokens, with weights, data, code and logs released.

DeepSeek-V2.5 is a model released on 5 September 2024 that merged DeepSeek's chat and code models into one, keeping both conversational and coding ability while improving safety scores.

Hunyuan Turbo is Tencent's large language model, released on Tencent Cloud in September 2024 and tuned for lower training and serving costs. It was succeeded in February 2025 by Hunyuan Turbo S, which cut API prices further.

Yi-Coder is a family of open-source code models from 01.AI, released in September 2024 in 1.5B and 9B sizes, each with base and chat versions, covering 52 programming languages with context up to 128K tokens.

o1-mini is a smaller, faster and cheaper reasoning model that OpenAI released on 12 September 2024 alongside o1-preview, trained with reinforcement learning to reason step by step and rated particularly effective at coding.

Aya Expanse is a family of multilingual open-weight language models released by Cohere For AI in October 2024 in 8B and 32B sizes, covering 23 languages.

Ministral 3B and Ministral 8B are two small Mistral AI language models released on 16 October 2024 for on-device and edge use, offering long context, low API prices and, for the 3B, performance Mistral claimed beat its earlier 7B model.

Anthropic's small, fast language model announced in October 2024 as the successor to Claude 3 Haiku, claimed to reach the intelligence of the previous flagship at Haiku-like speed and to be especially strong at coding.

A large language model released by iFlytek in October 2024 and upgraded in January 2025, trained and served on the domestic Feixing No.1 computing platform; iFlytek reported it leading several Chinese and English benchmarks against GPT-4 Turbo and GPT-4o.

SmolLM2 is Hugging Face's second family of small language models, released in November 2024 at 135M, 360M and 1.7B parameters, with open weights and a data-centric training recipe.

Tülu 3 is Ai2's open post-training recipe and model family, released in 8B and 70B sizes on top of Llama 3.1 and combining instruction tuning, preference tuning with on-policy data, and reinforcement learning with verifiable rewards.

OLMo 2 is the second generation of the Allen Institute for AI's fully open language model family, released in 7B and 13B sizes with weights, data, code and training recipes all public.

QwQ-32B-Preview is an experimental reasoning research model released by the Qwen team on 28 November 2024, with 32B parameters, a 32K-token context and open Apache 2.0 weights. It scored strongly on maths benchmarks but the team documented clear failure modes.

A version of OpenAI's o1 that spends more compute before answering, launched on 5 December 2024 inside the $200-a-month ChatGPT Pro plan and added to the API on 19 March 2025.

A family of three instruction-tuned Korean–English language models released by LG AI Research in December 2024, ranging from 2.4B to 32B parameters, with 32K-token context and weights out under a research-use licence.

Llama 3.3 70B is a text-only, instruction-tuned model Meta released on 6 December 2024. Meta said it performs close to the 405B Llama 3.1 at much lower cost and can run on a typical developer workstation.

A feature in the Gemini app that plans multi-step web research, gets the user's approval, browses for a few minutes and returns a report with links to its sources; later offered to developers as a managed agent in the Gemini API.

Project Mariner is a Google research prototype of a browser agent, announced with Gemini 2.0 in December 2024. Running as an experimental Chrome extension, it completes web tasks by understanding what is on the screen, and its computer-use capabilities later moved toward Google's developer APIs.

Falcon3 is TII's family of open-weight language models from 1B to 10B parameters, released in base and instruct form; the 10B instruct model appeared on Hugging Face in December 2024 with a 32K context and a custom licence.

An open-weights 3-billion-parameter language model from Infinigence AI, released on Hugging Face in December 2024 under Apache-2.0, built on a Llama-2 architecture with a 32K-token context for local deployment.

Megrez-3B-Omni is an open-source model from December 2024 that understands images, text and audio on end-side devices, pairing a 2.29B-parameter language backbone with SigLip and Whisper encoders for about 4B parameters in total.

Cohere North is an enterprise AI workspace that combines large language models, search and AI agents. It opened to early-access users in January 2025 and became generally available in August 2025.

Baichuan-M1 is Baichuan's medical model line, released on 24 January 2025 with Baichuan-M1-preview and the open Baichuan-M1-14B, which Baichuan described as trained from scratch for medicine.

Falcon Shores is an Intel chip that the company said in January 2025 it would use only as an internal test chip, never bringing it to market, while pointing to Jaguar Shores as its route into rack-scale AI data-centre systems.

Mistral Small 3 is a 24-billion-parameter open-weight language model from Mistral AI, released in January 2025 under Apache 2.0 and tuned for low-latency inference on a single GPU or laptop.

An experimental Google model released in February 2025 as part of the Gemini 2.0 family, described by Google as its best yet for coding and complex prompts, with a 2-million-token context window and tool use.

Phi-4-mini is a 3.8B-parameter text model and Phi-4-multimodal a 5.6B model that adds images and speech, both released by Microsoft in February 2025 as the small, efficiency-focused members of the Phi-4 family.

AI21 Labs' March 2025 hybrid SSM-Transformer instruction-following models, released as Jamba Large 1.6 and Jamba Mini 1.6 with 256k-token context, gated open weights and benchmarks AI21 compared against Mistral Large 2, Llama 3.3 70B and Command R+.

Cohere's first vision-language model, released in March 2025 as open-weight 8B and 32B versions. It handles images and text in 23 languages and, in Cohere's comparisons, beats several much larger models, under a non-commercial licence.

Reka Flash 3 is a 21-billion-parameter general-purpose reasoning model from Reka, trained from scratch and released with open weights under the Apache 2.0 licence on 11 March 2025.

The largest model in Ai2's OLMo 2 family: a 32-billion-parameter language model released with its data, code, weights and training details, and reported to outperform GPT-3.5-Turbo and GPT-4o mini on academic benchmarks.

Mistral Small 3.1 is an open-weight model released by Mistral AI in March 2025 that adds image understanding and a 128K-token context window to the Small line, runs on a single consumer GPU, and was later superseded by newer Small releases.

Llama Nemotron is NVIDIA's family of open reasoning models in Nano, Super and Ultra sizes, launched in March 2025 and built on Meta's Llama 3 models.

DeepSeek-V3-0324 is a March 2025 update to DeepSeek-V3 that keeps the same model structure while raising benchmark scores and moving to the MIT License.

xLAM-2 is Salesforce's family of large action models, from 1 billion to 70 billion parameters, trained on simulated agent–human data and tuned for multi-turn tool use rather than single-call function calling.

Llama 4 Behemoth is a very large Meta model previewed on 5 April 2025 as the teacher for the Llama 4 family. It was never publicly released.

ERNIE X1 Turbo is a Baidu reasoning model released on 25 April 2025, positioned as a cheaper and more capable successor to ERNIE X1, with stronger chain-of-thought, multimodal and tool-use abilities.

ERNIE 4.5 Turbo is Baidu's multimodal large language model released on 25 April 2025, presented as a faster successor to ERNIE 4.5 with lower hallucination rates and sharply cut API prices.

A 14-billion-parameter open-weight reasoning model released by Microsoft in April 2025, built by fine-tuning Phi-4 on reasoning traces generated by OpenAI's o3-mini and published under the MIT licence.

Amazon Nova Premier is the most capable model of the first Nova generation, available in Amazon Bedrock from 30 April 2025. It handles text, images and video with a 1-million-token context window, and also serves as a teacher model for smaller Nova models.

Mistral Medium 3 is a proprietary model released by Mistral AI on 7 May 2025, marketed as a mid-sized system that nears larger competitors at a much lower price and can be self-hosted on four GPUs.

Seed-Coder is a family of open code models released by ByteDance Seed in May 2025, with 8B-parameter Base, Instruct and Reasoning variants trained on data filtered by language models rather than hand-written rules.

Sarvam-M is Sarvam AI's open-weights 24-billion-parameter model, post-trained from Mistral Small for eleven Indian languages. It runs in think and non-think modes and reports large gains on Indic, maths and coding benchmarks.

An eight-terabyte collection of public domain and openly licensed text for pretraining language models, released together with two 7B-parameter models trained on it to show the data is enough.

MiniCPM4 is a language model built for end-side devices rather than servers, pairing a trainable sparse attention mechanism with a compact data recipe and an inference stack tuned for long-context speed.

Apple Foundation Models (2025) are the second generation of Apple's models. They add image understanding and a mixture-of-experts server design, and they open an on-device model to developers through the Foundation Models framework.

OpenAI o3-pro is a version of the o3 model released on 10 June 2025, which OpenAI described as its most capable model at the time and recommended for hard questions where reliability matters more than speed.

Magistral is Mistral AI's first family of reasoning models, launched in June 2025 with an open-weight Small model and an enterprise Medium model. It supports multilingual reasoning and was later updated and retired.

Gemini 2.5 Flash-Lite is the lowest-cost model in Google's Gemini 2.5 family, built for high-throughput work such as classification and summarisation, with thinking off by default and a controllable thinking budget.

Gemma 3n is a mobile-first open model in Google's Gemma family, released in June 2025 in two sizes that run in roughly 2–3 GB of memory and natively accept image, audio, video and text input.

Hunyuan-A13B is Tencent's open-weight mixture-of-experts language model, released on 27 June 2025 with 80 billion total parameters, 13 billion active per token and a 256K context.

Jamba 1.7 is a July 2025 update to the Jamba family, with Large and Mini versions offering 256k-token context, nine languages, and improved grounding and steerability.

SmolLM3 is a 3-billion-parameter open model released by Hugging Face on 8 July 2025. It supports six languages and up to 128K tokens of context, with switchable reasoning, under Apache 2.0.

Voxtral is Mistral AI's family of speech-understanding models, released on 15 July 2025 in 24B and 3B sizes under Apache 2.0; it transcribes long audio, answers questions about it, and turns spoken requests into function calls.

Megrez2-3x7B-A3B is a mixture-of-experts language model from Infinigence AI built for on-device deployment, in which groups of adjacent layers share one pool of experts so that it stores 7.5 billion parameters but activates only 3 billion per token.

Command A Vision is Cohere's first vision-language model in the Command family, released in July 2025. It pairs a 112-billion-parameter Command A language model with a SigLIP2 image encoder to read images, charts and documents and answer in text.

Baichuan-M2 is an open-source medical-enhanced model with 32 billion parameters, built on Qwen2.5-32B and released under Apache 2.0 for research and reference.

Claude Opus 4.1, released on 5 August 2025, upgraded Claude Opus 4 for agentic tasks, real-world coding and reasoning, at the same price. Anthropic later succeeded it with Claude Opus 4.5.

Cohere's first reasoning model, announced in August 2025: a 111-billion-parameter hybrid model that can switch its thinking on or off, built for agentic tasks and tool use, with open non-commercial weights.

Microsoft AI's first foundation model trained entirely in-house, announced on 28 August 2025. A mixture-of-experts text model, it opened for public testing on LMArena and was presented as one of several specialised models Microsoft intends to coordinate.

GLM-4.6 is an open-weights large language model released by Zhipu in late September 2025, with a 200K-token context window, about 357 billion parameters and an MIT licence, aimed at coding, reasoning and agent work.

Qwen3-Next is an open-weight model from Alibaba's Qwen team, released in September 2025 under Apache 2.0. It has 80B total parameters, only 3B active per token, and a hybrid attention design later used in Qwen3.5.

A 1.7-billion-parameter English model from Writer, fine-tuned from Qwen2.5-1.5B and released under Apache 2.0 with a 131,072-token context window, presented around grade-school mathematics.

A deep-thinking mixture-of-experts model from Baidu with 21 billion total and 3 billion active parameters, a 128K context window and an Apache 2.0 license, open-sourced in September 2025 alongside ERNIE X1.1.

ERNIE X1.1 is Baidu's reasoning model unveiled in September 2025, an update to ERNIE X1 that the company reports as stronger in factuality, instruction following and agentic tasks, built on ERNIE 4.5 and released alongside PaddlePaddle v3.2.

A 1.7-billion-parameter reasoning model released by Writer in September 2025, fine-tuned from Nvidia's OpenReasoning-Nemotron-1.5B, with strong competition-mathematics scores and an Apache 2.0 licence.

An OpenAI model released in September 2025 that adapted GPT-5 for agentic coding, served as the default model in Codex, and was shut down in July 2026.

xAI's low-cost reasoning model released on 19 September 2025, with a 2-million-token context window and far lower prices than Grok 4 at comparable benchmark scores.

An experimental DeepSeek model released on 29 September 2025 that introduced DeepSeek Sparse Attention to make long-context training and inference cheaper, shipped with open weights and an immediate API price cut.

OpenEnv is an open framework and community hub from Hugging Face for building reinforcement-learning environments as containerized services that expose an agent interface, MCP and the Gym training protocol. It was contributed to the Agentic AI Foundation in October 2026.

Tinker is a managed API from Thinking Machines Lab, announced on 1 October 2025, for fine-tuning language models while users choose the algorithms and data.

A 3-billion-parameter open reasoning model from AI21 Labs, released under the Apache 2.0 licence with a hybrid Transformer-Mamba design, a 256k-token context window and local execution on phones, Macs and PCs.

Claude Haiku 4.5, released on 15 October 2025 and available to all users, offered coding performance close to Sonnet 4 at one-third of the cost and more than twice the speed.

iFlytek's Spark X1.5 is a deep reasoning model released in November 2025 that runs on a domestic computing platform built with Huawei, using a 293-billion-parameter mixture-of-experts design.

Grok 4.1 is xAI's November 2025 update to Grok 4, aimed at emotional intelligence, creative writing and fewer hallucinations. Grok 4.1 Fast, a 2-million-token tool-calling model, followed on 19 November 2025.

An agentic coding model in OpenAI's GPT-5.1 family, released in November 2025. It was the first OpenAI model trained to work across multiple context windows and in Windows environments, and served as the default model in Codex until its retirement in 2026.

Amazon Nova 2 is the second generation of Amazon's Nova models, released at AWS re:Invent in December 2025. Its first member, Nova 2 Lite, is a fast reasoning model available in Amazon Bedrock with a 1-million-token context.

Three small dense models from Mistral AI — 3B, 8B and 14B — released in December 2025 under Apache 2.0, each in base, instruct and reasoning variants with image understanding.

NVIDIA Nemotron 3 is an open family of hybrid mixture-of-experts language models, announced in December 2025 in Nano, Super and Ultra sizes, with up to one million tokens of context.

Doubao Large Model 1.8 (Doubao-Seed-1.8) is ByteDance's Volcano Engine flagship released on 18 December 2025. It is built for multimodal agent tasks, with stronger tool calling and computer-operation abilities.

Baichuan-M3 is a new-generation open-source medical-enhanced language model with 235 billion parameters, built on Qwen3-235B-A22B and released under Apache 2.0 in January 2026.

Grok 4.20 is xAI's 2026 model family, released in reasoning, non-reasoning and multi-agent versions with a 1-million-token context window; Grok 4.3 followed with the same context and the same prices.

A large language model released by iFlytek on 11 February 2026, trained entirely on domestic computing power and aimed at practical use in education, healthcare, automotive and agents.

Gemini 3.1 Flash-Lite is a low-cost, low-latency multimodal model in Google's Gemini 3 family, released in preview in March 2026 and generally available from May 2026 for high-frequency, lightweight tasks.

Ai2's fully open 7B model, which mixes Transformer attention with Gated DeltaNet linear recurrent layers and matches its predecessor's MMLU accuracy using 49% fewer training tokens.

Mistral Small 4 is an open-weight mixture-of-experts model released in March 2026 that folds instruction-following, step-by-step reasoning and agentic coding into one model, with a switch that controls how much reasoning it does.

A 128B-parameter open-weights multimodal model from Mistral AI, listed in April 2026 for agentic and coding work, briefly the default in Le Chat before being superseded by Mistral Large 4.

ERNIE 5.1 is Baidu's large language model released on 9 May 2026, derived from ERNIE 5.0's pre-training at about 6% of the usual pre-training cost, and it placed at the top of Chinese models on LMArena text and search.[^1][^3]

Gemini Spark is Google's personal AI agent inside the Gemini app, announced at Google I/O on 19 May 2026. It runs around the clock and carries out tasks on the user's behalf, moving from trusted testers to the macOS app and to more than 160 countries.

Released on 28 May 2026 at the same price as Opus 4.7, Claude Opus 4.8 added a fast mode at 2.5 times the speed, with fewer unnoticed flaws in its own code. Anthropic called it a modest but tangible improvement.

Apple Foundation Models 3 (AFM 3), published on 8 June 2026, is a family of five models: two on-device models, including a sparse 20B multimodal model, and three cloud models, one of which is optimised for NVIDIA GPUs.

Grok 4.5 is an xAI model built for coding, agentic and knowledge work, announced on 16 July 2026. It was trained with Cursor on tens of thousands of NVIDIA GB300 GPUs and shipped as the default model in the Grok Build coding tool.

The fastest model in Google's Gemini 3.5 series, released in July 2026 for high-throughput agentic work, and slated to replace gemini-3.1-flash-lite and take over traffic from gemini-2.5-flash.

Gemini 3.6 Flash is the Gemini model Google released on 21 July 2026 and called its workhorse: better coding, knowledge work and multimodal results than 3.5 Flash, fewer output tokens, built-in computer use and tiered pricing.

Grok 4.6 is an xAI language model released on 12 August 2026, built on Grok 4.5 and aimed at long-running agents and interactive, visual work, with a 500,000-token context window and text and image input.

A Google model released on 13 August 2026 as its most intelligent workhorse for coding and agents, priced with a scheduled increase and superseded three weeks later by 3.8 Flash.

Granite 4.2 is IBM's open-weight model generation, released on 25 August 2026 in 3B, 8B and 30B dense sizes, with a reasoning-capable 8B model under the permissive Apache-2.0 licence.

iFlytek Spark X2.5 is a 293B-A30B mixture-of-experts language model released on 7 September 2026, trained and served entirely on domestic computing power, with a 256K context window and support for more than 200 languages.

Grok 4.7 is xAI's September 2026 flagship model for coding and knowledge work, built on a larger base model than Grok 4.6 with longer reinforcement learning and a new safeguard stack, at unchanged token prices.

PLaMo 3 Translate is a family of translation-specialised language models from Preferred Networks, released in four sizes from 0.8B to 31B parameters, covering 53 languages and shipping in several quantised formats.

ERNIE 5.5 Flash Base is Baidu's openly published base checkpoint from November 2026: a text-only sparse mixture-of-experts model with 280B total and 14B activated parameters, released for fine-tuning rather than as a finished assistant.

The Xicloud C600 is a MetaX data-center GPU built on the company's own GPGPU architecture, offering very large ECC-protected memory, MetaXLink scaling up to 128-GPU supernodes, a 1,000 W OAM 2.0 module, and air or liquid cooling.

Timothée Lacroix is a co-founder and the Chief Technology Officer of Mistral AI, and a co-author of the LLaMA paper and the Mistral 7B paper.

Guillaume Lample is a co-founder and Chief Science Officer of Mistral AI. He was the last of 14 authors on the LLaMA paper and co-authored the Mistral 7B paper.

Mark Chen is an OpenAI researcher who led the team behind DALL·E 2 and worked on Codex; in March 2025 he moved into an expanded Chief Research Officer role.

Lilian Weng is an AI researcher whose main public writing is her blog Lil'Log, which has carried her learning notes since 2017. Chinese-language coverage describes her as a Peking University alumna and reports on her years at OpenAI and her later moves.

AI researcher who was first author of the 2016 paper introducing neural architecture search and a co-author of Switch Transformers; his X profile lists him as CTO and co-founder of Thinking Machines Lab and VP Research at Google DeepMind.

GroqChip is Groq's AI processor, built for fully deterministic execution with no run-to-run variation. It pairs a 14-nanometer die with 230 MB of on-die SRAM, up to 750 INT8 TOPS and RealScale chip-to-chip links, and ships both as a chip and on the PCIe GroqCard.

SambaNova SN40L is an AI chip that SambaNova calls a Reconfigurable Dataflow Unit, pairing dataflow-style parallel execution with a three-tier memory system spanning on-chip SRAM, co-packaged HBM and DDR DRAM.

Wu Yonghui is an AI researcher who spent nearly two decades at Google, worked on Google Brain and became a vice president of research at Google DeepMind, before joining ByteDance's Seed large-model team in 2025.

Jingren Zhou is a computer scientist at Alibaba, where he is chief technology officer of Alibaba Cloud Intelligence, leads the Tongyi AI lab and joined the Alibaba Partnership in 2025; he was named a 2024 ACM Fellow.

Yan Junjie is a co-founder and chief executive officer of the Chinese AI company MiniMax, which he led to a Hong Kong listing on 9 January 2026; he was previously an executive at SenseTime.

Chinese AI researcher and entrepreneur who co-founded Moonshot AI in Beijing in 2023 and runs it as CEO; the company released the Kimi assistant and the open-weight Kimi K2 and K3 models.

Agent memory is the part of an agent system that stores information beyond the current context window so it can be recalled later. MemGPT (2023) framed the limited context window as the core constraint such designs must work around.

LM Studio is a closed-source desktop app for macOS, Windows and Linux that lets people download, run and chat with large language models entirely on their own computer, and serves them to other software through OpenAI-compatible local endpoints.

Sam McCandlish is a co-founder and chief technology officer of Anthropic, a co-author of influential work on scaling laws and GPT-3, and the company's first Responsible Scaling Officer.

AI researcher and research administrator who worked at IBM's Watson Group and at JD.com, and has been director and chief scientist of the Shanghai AI Laboratory since April 2024.

Zhiyuan Liu is a tenured associate professor at Tsinghua University and co-founder and chief scientist of Beijing ModelBest, a company working to run capable large models on local devices.

Mike Krieger, co-founder and CTO of Instagram, joined Anthropic as Chief Product Officer on 15 May 2024, taking charge of product engineering, product management and product design.

Brad Lightcap is an OpenAI executive who served as chief financial officer and then chief operating officer, later overseeing business strategy, partnerships, infrastructure and global deployment before leaving the company in 2026.

AI researcher and entrepreneur who held research positions at OpenAI, Google and DeepMind before co-founding the AI search company Perplexity in 2022, where he serves as CEO.

Daniel Gross co-founded Safe Superintelligence Inc. with Ilya Sutskever and Daniel Levy and served as its chief executive until early July 2025, when he moved to Meta's newly launched Meta Superintelligence Labs.

Zhang Peng is chief executive officer of Zhipu AI (Z.ai). He led the company through its Hong Kong listing on 8 January 2026, described as the first by a major Chinese LLM company, and through the GLM-4.5 to GLM-5.3 model releases.

Tang Jie is a Chinese computer scientist listed as a founder of Zhipu AI and its chief scientist, and a Tsinghua professor who created the AMiner academic database and co-authored the GLM pretraining work behind Zhipu's models.

Jacob Devlin is a research scientist and first author of BERT, a language model Google introduced in 2018 that reads text in both directions. He earlier worked on neural machine translation at Microsoft Research.

Jiang Daxin is founder and CEO of the Chinese AI company StepFun and a former Microsoft global vice president. He argues that scaling laws set the ceiling on model intelligence and that unifying multimodal understanding and generation is the path to AGI.

Ori Goshen is an Israeli entrepreneur who co-founded AI21 Labs in November 2017 with Yoav Shoham and Amnon Shashua and serves as its chief executive.

Jack Clark is a co-founder of Anthropic, where he has led policy and later public-benefit work and, from March 2026, the Anthropic Institute. He is also a co-author of the GPT-3 paper.

Tom Brown is an AI researcher listed among the co-founders of Anthropic and the first author of the 2020 paper that introduced GPT-3, one of the most widely cited demonstrations of few-shot learning in large language models.

Connor Leahy is a German-American AI researcher and AI-safety advocate. He co-founded EleutherAI in 2020, co-founded the AI-safety company Conjecture, and in 2026 became US Director of the advocacy group ControlAI.

Wang Haifeng is Baidu's chief technology officer, an ACL and IEEE Fellow, and the first-listed author of the ERNIE technical reports behind the company's foundation models.

Yuewen is StepFun's AI assistant app, fully opened in March 2024, which reads text, data and charts in images and handles content creation, reasoning and analysis. In December 2024 its team absorbed the Maopaoya team, and the product was later renamed Step AI.

Wanzhi is a free bilingual AI productivity workspace released by 01.AI in May 2024 for the web and as a WeChat mini program; the name later extended to an enterprise large-model platform.

Baixiaoying is Baichuan's first AI assistant app, launched on 22 May 2024 and powered by its Baichuan 4 model. It is built around search, including multi-round search, and document reading, and Baichuan has framed it as a step towards an AI "super app".

NVLM is a family of multimodal large language models from NVIDIA, described in a paper posted to arXiv in September 2024; its flagship NVLM-D-72B was later released as open weights for non-commercial use.

NVIDIA's Llama-3.1-Nemotron-70B-Instruct is a version of Meta's Llama-3.1-70B-Instruct tuned with reinforcement learning from human feedback to give more helpful answers; on release it topped three automatic alignment benchmarks.

Nemotron Nano 2 is NVIDIA's 9B reasoning model, built on a hybrid Mamba-Transformer design and released with open weights and most of its training data; NVIDIA reports much higher throughput than similar-sized models.

Blackhole is Tenstorrent's AI accelerator chip; Galaxy Blackhole is a server holding 32 of them, rated at 23 PFLOPS of block FP8, with Ethernet links wiring chips and clusters together.