Total 671 · Clear filters
On 26 February 2024, Mistral AI's API changelog announced Le Chat, a beta chat interface for talking to its models directly, released alongside the Mistral Large and Mistral Small models and new function calling and JSON mode features.
On 26 February 2024 Mistral AI released Mistral Large and Mistral Small, and announced a multi-year partnership with Microsoft that made Mistral Large available first on Azure AI. Le Chat, its assistant, launched in beta the same day.
On 22 February 2024 Stability AI announced an early preview of Stable Diffusion 3, a text-to-image suite from 800M to 8B parameters that combines a diffusion transformer with flow matching.
On 21 February 2024 Google released Gemma, a family of open models in 2B and 7B sizes, with pre-trained and instruction-tuned versions, weights distributed through Kaggle and Hugging Face and terms permitting commercial use by organisations of any size.
Announced on 15 February 2024, Gemini 1.5 Pro was a mixture-of-experts model offering a 128,000-token context window by default and up to 1 million tokens in private preview — far more material taken in at once.
On 15 February 2024 OpenAI previewed Sora, a text-to-video model that it said could generate clips up to one minute long; the model reached ChatGPT Plus and Pro users in the US and Canada on 9 December 2024.
On 8 February 2024 SenseTime announced SenseNova 4.0, including SenseChat V4 in 4K, 32K and 128K context versions, specialised models for data analysis, medicine and vision, a tool-calling API, and a HumanEval score above GPT-4's.
On 4 and 5 February 2024 Alibaba released the Qwen1.5 open model family, with base and chat models from 0.5B to 72B parameters and a 32K context window across the lineup.
On 1 February 2024, the Allen Institute for AI released OLMo, a language model released together with its pretraining data, training code, logs and more than 500 checkpoints per model, all under the Apache 2.0 licence.
In February 2024 Cohere For AI released Aya, an instruction-tuned model covering 101 languages. It has 13 billion parameters, uses an Apache 2.0 licence and beats mT0 and BLOOMZ on most tasks, according to its authors.
On 30 January 2024 iFlytek released Spark V3.5, which it described as the first Spark model trained on the fully domestic FlyingStar One platform and which it compared with GPT-4 Turbo across several benchmarks.
On 29 January 2024 Baichuan released Baichuan 3, a closed model with more than one hundred billion parameters, and reported its own benchmark results against GPT-4.
On 23 January 2024, 01.AI released the open vision-language models Yi-VL-34B and Yi-VL-6B. Yi-VL-34B ranked first among open models on the MMMU and CMMMU benchmarks.
On 10 January 2024 OpenAI opened the GPT Store, a marketplace for custom versions of ChatGPT. More than three million GPTs had already been built by the time it launched, after the opening slipped from its originally planned date.
On 9 January 2024 DeepSeek released DeepSeekMoE, a mixture-of-experts architecture built around fine-grained experts and shared experts, with an open-weights 16B model.
On 21 December 2023 Midjourney released version 6 of its image model in alpha, trained from scratch over nine months and better at rendering text. V6.1 followed on 30 July 2024 with faster, more coherent output and became the default model.
On 12 December 2023 Microsoft Research announced Phi-2, a 2.7-billion-parameter model it pitched as a playground for researchers, released under the MIT licence.
On 11 December 2023 Mistral AI released Mixtral 8x7B, an open-weight mixture-of-experts model that activates only part of its parameters for each word. It aimed to pair a large total capacity with lower compute per token.
On 6 December 2023 Google announced Gemini 1.0, a model family built to learn text, images, audio and video together from the start. Google called its largest version its most capable model yet.
On 29 November 2023 DeepSeek released DeepSeek LLM in 7B and 67B sizes, each with base and chat versions, trained on 2 trillion tokens.
On 29 November 2023, during AWS re:Invent, Amazon made the Titan Text Express and Titan Text Lite models generally available in Amazon Bedrock and previewed Titan Image Generator, which adds an invisible watermark to every image.
