fresh out the oven
every AI model and tool launch we've logged, newest first, each checked against the maker's own post. 18 so far.
Muse Spark Open
Meta: open weights on Hugging Face: 816B-param MoE (42B active), 1M-token context, text, image, audio and video in. built for coding and agents.
primary source ↗Claude Opus 5.5 fast mode
Anthropic: out of research preview, waitlist gone: same Opus 5.5, up to 2.5x faster, $8/$40 per M tokens on the Claude Platform.
primary source ↗GPT-Live-1.1
OpenAI: voice update in ChatGPT, Codex and dots: new voices, better accents, less waiting. rolling out to all users; API coming soon.
primary source ↗Step 5 Preview (open weights)
StepFun: 600B MoE (27B active), 1M-token context, text and vision input; BF16 and FP8 weights on Hugging Face and ModelScope.
primary source ↗Nemotron 3.5 Super VL
NVIDIA: open 120B model for always-on agents, on Hugging Face and as NVIDIA NIM; Nemotron 3.5 Bolt 74B-A9B for local agents ships alongside.
primary source ↗Grok Voice 3.0
SpaceXSI: speech-to-speech model for voice agents, replies in under a second; in the API at $0.08/min, same price as 2.0.
primary source ↗Grok Build
SpaceXSI: terminal coding agent with TUI, headless mode and ACP editor integration; pricing not yet published.
primary source ↗Chainguard Workspaces (beta)
Chainguard: hardened microVMs for coding agents: credential isolation, optional egress controls, signed run attestations.
primary source ↗Gemini 4 Argon (GA)
Google: frontier reasoning model, 1M-token limit; $2/$10 per M tokens intro pricing until nov 30.
primary source ↗Muse Wonder & Muse Wonder Full
Meta: agentic frontier models via the Meta Model API and Muse Code.
primary source ↗Clef-omni
Cloudflare: open-weight multimodal decision model (text, image, audio, video) on Workers AI; $0.15/M input tokens.
primary source ↗Genkit Dart 1.0
Google: stable release of Google's open-source framework for AI features and agents in Dart/Flutter.
primary source ↗Mellum2.1
JetBrains: open-weight 12B MoE (2.5B active) coding-agent model, Apache 2.0; RL-trained for repo explore/edit/self-check.
primary source ↗GPT-6 + Intelligent UI in ChatGPT
OpenAI: GPT-6 rolls out to all ChatGPT tiers with interactive UI answers.
primary source ↗Claude Haiku 5.5
Anthropic: cheapest, fastest Claude; $0.10/$0.50 per M tokens; Sonnet 5.5 cache reads halved.
primary source ↗Mistral Large 4 (preview)
Mistral: 1T-param MoE (52B active); preview API now, open weights promised end of october.
primary source ↗Nano Banana 2.1
Google: image model update: better text rendering, mask editing, consistency; GA on Gemini API.
primary source ↗Amazon Nova 2.5 Sonic
Amazon: speech-to-speech model for voice agents, GA on Bedrock.
primary source ↗
see the rankings →