Total 671 · Clear filters
In January 2025 MiniMax released the open-weight MiniMax-01 series, a language model and a vision-language model with a 4-million-token inference context. Weights were posted to Hugging Face on 12 January 2025.
In January 2025 Alibaba released Qwen2.5-VL in four sizes. The model works as a visual agent, follows videos longer than an hour and returns structured coordinates and document data.
ByteDance's AI coding tool TRAE launched in January 2025. By the end of 2025 it had 6 million registered users, and its open-source Trae Agent reported 75.20% on SWE-bench Verified, first on the leaderboard.
On 26 December 2024 DeepSeek released DeepSeek-V3, a mixture-of-experts model with 671 billion total parameters, together with its open model weights, a technical report and code.
In December 2024 Infinigence AI released Megrez-3B, a two-model family — Instruct and Omni — built for phones and other end-side hardware, with open weights under Apache-2.0.
On 16 December 2024 Google announced Veo 2, a video model it said could produce clips at up to 4K resolution, and put it in its VideoFX experiment; it reached the Gemini app in April 2025.
On 14 December 2024, TII released the Falcon3 family of open language models in 1B, 3B, 7B and 10B sizes, each in base and instruct form, with the 10B built by depth up-scaling the 7B rather than designed from scratch.
On 13 December 2024 DeepSeek released the DeepSeek-VL2 vision-language models, a mixture-of-experts series in three sizes, with dynamic tiling for high-resolution images.
On 12 December 2024 Microsoft released Phi-4, a 14-billion-parameter model specialised in complex reasoning such as maths, with open weights following on 8 January 2025.
On 11 December 2024 Google announced Gemini 2.0, a model line pitched at the "agentic era", with Gemini 2.0 Flash released first to developers and app users.
On 11 December 2024, Google added Deep Research to Gemini Advanced: an assistant that draws up a research plan, browses the web for a few minutes and writes a sourced report. It launched the same day Google introduced Gemini 2.0 Flash Experimental.
On 6 December 2024 LG AI Research published the technical report for EXAONE 3.5, a three-size family (2.4B, 7.8B, 32B) with 32K context and research-only licensing, whose weights had appeared on Hugging Face five days earlier.
On 6 December 2024 Meta released Llama 3.3, a 70B text model that Meta said performs similarly to Llama 3.1 405B at much lower cost. It was the last Llama 3 release before Llama 4.
On 5 December 2024, OpenAI launched ChatGPT Pro, a $200-a-month plan with unlimited access to o1, o1-mini, GPT-4o, Advanced Voice and o1 pro mode; full o1 also arrived in ChatGPT the same day.
Genie 2 is a Google DeepMind world model, announced on 4 December 2024, that turns a single image into a playable, action-controllable 3D world that people or AI agents can explore with a keyboard and mouse.
On 3 December 2024 Tencent released the inference code and weights of HunyuanVideo, an open-source video generation model with over 13 billion parameters.
On 3 December 2024, at AWS re:Invent, Amazon introduced the Nova family of foundation models, starting with Nova Micro, Lite and Pro for understanding tasks and Nova Canvas and Reel for images and video. Nova Premier was announced for early 2025.
In December 2024 Kuaishou released Kling 1.6, an upgrade to its video generation model that improved how faithfully clips follow prompts about motion and camera movement, and raised visual quality.
On 28 November 2024 the Qwen team released QwQ-32B-Preview, an experimental open-weights reasoning model with a 32K-token context that scored strongly on maths and coding benchmarks while showing clear limitations.
On 26 November 2024 Ai2 released OLMo 2 in 7B and 13B sizes, publishing data, code, recipes and checkpoints alongside the weights, and described the pair as the best fully open language models to date.
On 25 November 2024 Anthropic open-sourced the Model Context Protocol, a standard for connecting AI assistants to the systems where data lives. In December 2025 Anthropic donated it to the Agentic AI Foundation under the Linux Foundation.
