Mistral AI releases Devstral 2 and the Mistral Vibe CLI
On 9 December 2025 Mistral AI released Devstral 2, a 123B dense coding model, and the smaller Devstral Small 2, alongside Mistral Vibe, an open-source command-line coding agent built on them.
Total 1,098
Moore Threads and MetaX list on the STAR Market
In December 2025, Chinese GPU makers Moore Threads and MetaX listed on the STAR Market, with first-day gains that drew attention to domestic AI chip companies raising public capital.
Tencent releases HY 2.0 with Think and Instruct versions
On 5 December 2025 Tencent released HY 2.0 in Think and Instruct versions, a 406-billion-parameter mixture-of-experts model with a 256K context, integrated into Yuanbao and ima.
Kuaishou releases Kling 2.6 with native audio
On 3 December 2025 Kling AI released Video 2.6, a video model that generates picture and sound together — speech, singing and ambience — with Chinese and English voices in clips of up to 10 seconds.
Claude Code reaches $1B run-rate; Anthropic acquires Bun
On 3 December 2025 Anthropic said Claude Code had reached $1 billion in run-rate revenue, and agreed to acquire Bun, a JavaScript runtime that had already worked with Anthropic on Claude Code's native installer.
Mistral AI releases the Mistral 3 family, including Mistral Large 3
On 2 December 2025 Mistral AI released the Mistral 3 family: Mistral Large 3, a 675B mixture-of-experts model, and three Ministral 3 models, all under Apache 2.0.
Amazon launches Nova 2 and Nova Forge at re:Invent 2025
On 2 December 2025, at AWS re:Invent, Amazon released the Nova 2 family, led by the reasoning model Nova 2 Lite. The same event introduced Nova Forge, made Nova Act generally available and launched Trainium3 UltraServers.
DeepSeek releases V3.2 and gold-medal-level V3.2-Speciale
On 1 December 2025 DeepSeek released V3.2, the official successor to V3.2-Exp, and V3.2-Speciale, a version built to maximize reasoning. Both had their weights published on Hugging Face, and V3.2-Speciale was reported to reach gold-level results at several international competitions.
Kuaishou launches Kling O1 unified multimodal video model
On 1 December 2025 Kuaishou's Kling AI released Kling O1, which the company described as the world's first unified multimodal video model, combining video generation and editing from text, image, video and subject inputs.
Hugging Face announces Transformers v5
On 1 December 2025 Hugging Face released Transformers v5.0.0rc0, the first major version since v4 in November 2020. The library grew from about 40 architectures to more than 400.
Runway releases Gen-4.5
On 1 December 2025 Runway released Gen-4.5, a text-to-video model that reached the top of the Artificial Analysis leaderboard at 1,247 Elo, with claimed gains in physics, scene composition, character expression and controllability.
Black Forest Labs raises $300 million at a $3.25 billion valuation
On 1 December 2025 the German image-model developer Black Forest Labs said it had raised a $300 million Series B co-led by Salesforce Ventures and Anjney Midha's AMP, valuing the company at $3.25 billion.
MiniMax releases MiniMax-M2.1 for multilingual coding
MiniMax released MiniMax-M2.1, an open-weights model built for multilingual coding and tool use, with SWE-bench Multilingual rising from 56.5 to 72.5 over its predecessor M2.
Zhipu releases GLM-4.7 for agentic coding
In December 2025 Zhipu released GLM-4.7, an MIT-licensed open model aimed at multilingual agentic coding, with higher scores than GLM-4.6 on SWE-bench and Terminal Bench 2.0.
DeepSeekMath-V2 reaches gold-medal level on IMO 2025 problems
On 27 November 2025 DeepSeek released DeepSeekMath-V2, a model trained to check its own proofs. DeepSeek reported gold-level scores on IMO 2025 problems and 118 of 120 points on Putnam 2024, the latter with scaled test-time compute.
MCP publishes the November 2025 specification
On 25 November 2025 the Model Context Protocol project published its November 2025 specification, adding task-based workflows, simpler authorisation, security and enterprise features, a formal extension mechanism, and new elicitation and sampling modes.
Black Forest Labs releases FLUX.2
On 25 November 2025 Black Forest Labs released the FLUX.2 image models, with hosted pro and flex variants and the open-weight 32B dev model. The smaller klein models followed in January 2026.
Anthropic releases Claude Opus 4.5
On 24 November 2025 Anthropic released Claude Opus 4.5 at $5 and $25 per million tokens, a third of the previous Opus price, with a new effort parameter for trading capability against cost.
Google launches Nano Banana Pro
On 20 November 2025 Google released Nano Banana Pro, an image model built on Gemini 3 Pro that emphasises correctly rendered text in multiple languages, Search-grounded infographics and 2K/4K output.
Ai2 releases Olmo 3 with open reasoning models
On 20 November 2025 Ai2 released Olmo 3 in 7B and 32B sizes, in Base, Think, Instruct and RL Zero variants. Ai2 called Olmo 3-Think 32B the strongest fully open thinking model.
Commission proposes a digital omnibus that delays high-risk AI rules
On 19 November 2025 the European Commission proposed a digital omnibus on AI that would tie high-risk AI obligations to the availability of standards, stretch their start to at most 16 months, add a 2028 sandbox and strengthen the AI Office.