AI TrackerModels news,
for builders.
The latest models news from every AI provider: launches, price changes and retirements, summarized with what they mean for you.
Copilot CLI 1.0.106 adds GPT-6.1 Terra, a /code command and a Fusion tier setting
GitHub's Copilot CLI 1.0.106 adds:Thinking Machines releases Inkling 2 Small, a proprietary model with open weights promised
Thinking Machines introduced Inkling 2 Small, saying it scores 31 on the Artificial Analysis Intelligence Index, a big jump over Inkling Small. It is available for fine-tuning and inference on Tinker, its platform, and in Tinker Playground today. Thinking Machines says the full weights are planned soon; for now the model is proprietary.Arena ranks Tencent's Hy4 the best Chinese model in Agent Arena
Arena says Tencent's Hy4 is now the #1 Chinese model in its Agent Arena, and that this makes Tencent Hunyuan the only Chinese lab in the top 5. Arena bases the ranking on more than 12,000 real-world, long-horizon agentic sessions from its user community, up from about 8,500 in its earlier result, when it placed Hy4 #14 overall.World Labs releases Frameboy, a real-time demo of its Atlas world model
World Labs released Frameboy, a demo for exploring generated worlds in real time. It runs a highly optimized version of Atlas, the autoregressive world model with camera control that the company introduced two months ago, and streams from cloud GPUs at up to 15 frames per second. World Labs describes Atlas's camera control as pixel-perfect; that is the company's claim.OpenRouter's catalog gives two Meta Llama 3.2 models an expiration date of November 15
OpenRouter's model catalog now shows an expiration date of November 15, 2026 for two Meta models listed since September 2024: Llama 3.2 3B Instruct (meta-llama/llama-3.2-3b-instruct, 131,072-token context) and Llama 3.2 1B Instruct (meta-llama/llama-3.2-1b-instruct, 60,000 tokens). Neither entry had an expiration date before. This is a catalog field only; the entries do not say why or what happens to the models after that date.Artificial Analysis ranks FLUX 3 Image Klein #2 among open-weight text-to-image models
Artificial Analysis says Black Forest Labs' FLUX 3 Image Klein, a 5.6B-parameter open-weight image model released on October 15, is the #2 open-weight model on its AA-Image-T2I v2.1 leaderboard, neck and neck with #1 Qwen-Image-2.1. On AA-Image-Editing v2.1 it ranks #3 among open-weight models. The rankings are Artificial Analysis's own.Moonshot releases the open weights of Kimi K3.1 Pro
Moonshot posted on X that it is releasing the model weights of Kimi K3.1 Pro, on Hugging Face as moonshotai/Kimi-K3.1-Pro, and said "we'll keep advancing open intelligence." It had promised the weights by November 4 when it launched the model on October 23.Perplexity cuts Agent API prices at every effort level and adds an "ultra" level
Perplexity says every effort level of its Agent API now costs less with the same or higher research quality, and adds a new ultra level for the most demanding, long-running requests. Effort levels are bundled agent configurations, optimized for specific tasks; developers choose among fast, low, medium, high, xhigh and ultra.Gemini Omni 1.2 Flash is now live on Vercel's AI Gateway
Vercel says Google DeepMind's Gemini Omni 1.2 Flash video model is live on its AI Gateway, under the model id google/gemini-omni-1.2-flash. Vercel lists clips of 3 to 15 seconds with native audio, edits with up to six reference images, extensions to 30 seconds in total, and resolutions from 360p to 4K in landscape or portrait.Kimi K3.1 Pro is now available in Devin Desktop and Devin CLI
Cognition's Devin team posted that Moonshot's Kimi K3.1 Pro can now be used in Devin Desktop and Devin CLI.Arena ranks Gemini Omni 1.2 Flash #1 in text-to-video and image-to-video
Arena says Google's Gemini Omni 1.2 Flash debuts at #1 in both its Text-to-Video and Image-to-Video leaderboards and at #3 in Video Edit.OpenAI will retire gpt-5.5-cyber-preview on February 1, 2027
OpenAI's API deprecations page lists gpt-5.5-cyber-preview for shutdown on February 1, 2027, in an entry dated November 5. As the replacement it names "the most capable cyber model available to you", without naming a model.Design Arena ranks Ling-3.2-flash #10 on Image to Frontend Arena
Design Arena says Ant Ling's Ling-3.2-flash placed #10 on its Image to Frontend Arena with an Elo of 1238, among models from Meta, Anthropic and DeepSeek. It says the model sets a new Pareto frontier of preference against price, at $0.30 input and $0.90 output per 1M tokens, the cheapest for any model at that level of performance.Gemini Omni 1.2 Flash is now live on OpenRouter
OpenRouter says Google's Gemini Omni 1.2 Flash video model is live on its platform. It lists 15-second clips, resolutions from 360p to 4K, native audio, and about 30% lower per-second pricing than Omni 1.1. OpenRouter says the price is the same as going direct to Google, and that one API now covers video and chat.METR publishes its assessment of whether AI is accelerating AI development at Anthropic
METR, the AI evaluation nonprofit, published a report on a risk assessment in which a team within METR has worked since August with internal Anthropic information to judge whether AI itself is accelerating AI development. It also published a separate review of the AI R&D section of Anthropic's August 2026 risk report.Ant Ling introduces Ling-3.2-flash, a 560B multimodal model that it says beats Opus 5 on three benchmarks
Ant Ling announced Ling-3.2-flash, which it calls a native multimodal model with 560B total and 25B active parameters. It says the model is stronger at complex visual understanding, long-horizon multimodal agents, 3D understanding and interaction, and medical intelligence.Artificial Analysis scores open-weight Nanbeige4.3-33B 20 on its Intelligence Index, level with Qwen3.8 27B
Artificial Analysis says Nanbeige4.3-33B, the open-weight model from Nanbeige (the foundation-model team at the Chinese hiring platform BOSS Zhipin), scores 20 on its Intelligence Index v5, the same as Qwen3.8 27B (xhigh). Both are dense models, with 33B and 27B total parameters. Artificial Analysis says Nanbeige's model has stronger terminal use but weaker knowledge, and sets a new Pareto-optimal point for Intelligence Index against total parameters.Gemini Omni 1.2 Flash is not announced, but several accounts and OpenRouter's catalog show it
Google has not announced Gemini Omni 1.2 Flash, a video-generation model that would follow Omni 1.1 Flash. OpenRouter's model catalog also added a google/gemini-omni-1.2-flash entry on November 5, listing a 131,072-token context. Several accounts report seeing the model:OpenAI introduces GPT-6 Cyber for security research, for select Daybreak Red customers
OpenAI introduces GPT-6 Cyber, a security-research model that it says pairs stronger capabilities with better alignment. It says this is the model behind the Linux kernel research it announced the day before, and that security teams at Mozilla, Palo Alto Networks, Calif and Cisco, as well as OpenAI's own, are already using it.Google releases Gemini Omni 1.2 Flash with 15-second video and 30% lower pricing
Google says Gemini Omni 1.2 Flash is generally available on the Gemini API and Google Cloud, and that this is the first time Gemini Omni is generally available. It generates and edits video natively up to 15 seconds long (up from 10 seconds with Omni 1.1), with duration settable in one-second steps from 3 to 15 seconds.