Daily AI news, sourcedUpdates / Models
Updates Models · 505 items · newest issue first · page 12 of 26
AreaAllModelsAgents & codingApps & productsChips & infraResearchPricing & limitsBusiness & fundingPolicy & safetyOther
- vegaml open-sources Vega, an 800M/4B physics-based typed decision model with 73k context and images
- mlx-community publishes an fp32 MLX port of Big-LaMa for Apple Silicon inpaintingUnconfirmedHugging Face @mlx-communityOct 10Models
- webAI releases TwIL-LM3-Pro, a 3.66B formal-logic model built on Granite 4.2 3B
- Community run claims Qwen3.6-35B-A3B Q4_K_M hits ~600 tok/s prefill on a 6 GB RTX 2060Unconfirmedr/LocalLLaMAOct 10Models
- Nace.AI open-sources Drex 1.5, a 9B decision model that scores options in one pass, per MarkTechPostReportedMarkTechPostOct 10Models
- Community Qwen3.8-27B Heretic quant with MTP fits a 16 GB GPU at a claimed 55-68 tok/sUnconfirmedr/LocalLLaMAOct 10Models
- User benchmarks Qwen Flash-Next at 137 tok/s with 512K context on one RTX 5090, comparing Strata and InfernixUnconfirmedr/LocalLLaMAOct 10Models
- Infron adds seven low-refusal models for authorized red-teaming and security research
- Google deprecates Gemini 3.7 Flash in the Gemini API
- Alibaba Token Hub shows how Qwen3.8-Max and Qwen-Audio-3.1-Realtime move AI from chat to actionOfficialX @alibaba_cloudOct 10Models
- Interconnects author @natolambert to speak at Fireworks Forge on how open models win
- Apsara 2026 panel discusses moving open-source AI from benchmarks into enterprise production
- Mynet CTO says Qwen captures the structure and nuance of Japanese writing at Apsara 2026
- StepFun says Step 5 Preview is on OpenRouter's New & Trending leaderboardOfficialX @StepFun_aiOct 10Models
- Microsoft introduces Microsoft-Decision-1, a model for fast structured decision-makingOfficialX @satyanadellaOct 10Models
- MLX community publishes an 8-bit MLX conversion of the Clef Omni modelUnconfirmedHugging Face @mlx-communityOct 10Models
- Google employees are testing an internal Gemini 4 version called Carbon that feels like Opus 5.5 for coding, per Business InsiderReportedBusiness InsiderOct 10Models
- Talus, a 23M-parameter diffusion model for game terrain, runs in the browser on WebGPUUnconfirmedr/MachineLearningOct 10Models
- H2O.ai releases H2O-Lightning-4B, an Apache-2.0 decision model it says tops JevBench among open modelsUnconfirmedr/LocalLLaMAOct 10Models
- Mistral releases Voxtral Mini 4B Realtime Arabic for streaming speech-to-text across Arabic dialectsOfficialHugging Face @mistralaiOct 10Models