Daily AI news, sourcedUpdates / Models
Updates Models · 505 items · newest issue first · page 3 of 26
AreaAllModelsAgents & codingApps & productsChips & infraResearchPricing & limitsBusiness & fundingPolicy & safetyOther
- StepFun's Step 5 Preview goes live on DeepInfra at $1 input and $2.70 output per million tokens
- OpenAI upgrades GPT-6 Astra Pro in ChatGPT with faster responses and parallel agentsOfficialOpenAIOct 17Models
- Perplexity API changelog: pplx-decider-v1.2 and a new 4B pplx-decider-flash-v1.2 come to the Decisions API
- Intel CEO Lip-Bu Tan says Chinese frontier AI models are cheaper per token and being adopted 'very fast', per Reuters
- StarSkirmish Bench v0.2 ranks GPT-6 Astra first at writing StarCraft bots in a one-hour coding harness
- Tencent open-weights DSpine, a parallel speculative-decoding drafter for Qwen3-8B
- LocalLLaMA user says a LoRA fine-tune of d3-edge 0.6B beat d3 27B on typed decisionsUnconfirmedr/LocalLLaMAOct 17Models
- AlexWortega open-weights Qwen3.8-27B-antislop, RL'd against slop, ranking #4 on EQ-Bench 4OfficialAlexWortega on Hugging FaceOct 17Models
- LocalLLaMA user turns a 2D platformer LLM challenge into a public decision-making benchmark with a leaderboardUnconfirmedr/LocalLLaMAOct 17Models
- Vercel data shows TypeSafe AI's Jev decision model reached 35% of paid teams in two weeks, per TechCrunch
- Elorian launches Glimpse, a visual-reasoning model with a mid-reasoning crop-and-measure scratch space, per TechCrunchReportedTechCrunchOct 17Models
- OpenMOSS open-weights MOSS-Image, a 5B text-to-image family with native JSON layout controlOfficialOpenMOSSOct 16Models
- Hugging Face Transformers 5.20.1 adds Reflection AI Beam support and fixes older-torch loading
- Tencent Hunyuan releases Hy-Memory-30B-A3B, a ~3B-active long-term memory writer for Mem0, MemOS and Hy-MemoryOfficialTencent HunyuanOct 16Models
- NVIDIA releases a DFlash 2 speculative-decoding draft model for Meta's Muse Spark Open 816B
- Apollo Research finds GLM-5.3 sometimes reasons, unprompted, about how its actions could shape its own trainingOfficialApollo ResearchOct 16Models
- Amazon releases ALoDLM-1.7B and 8B, adaptively looped diffusion language models built on Qwen3
- Qwen releases BlenderWorld, a 223-task benchmark for visual agents building 3D scenes in Blender from real-world referencesOfficialQwenOct 16Models
- fal opens fal Worlds, a page of real-time generative worlds you can walk through and directOfficialfalOct 16Models
- Shanghai AI Lab makes Atria Dawn available through its official API, a 744B MoE model built on GLM-5.2