The week in AI
8 new models from 6 labs, 56 tool releases and 24 stories.
Download the card · This week’s link
New models · 8
All models- Gemini 3.8 Live Extended Thinking 15 Sep
- Claude Mythos 5.1 15 Sep
- Eleven Music v2.5 14 Sep
- Qwen3.8-LiveTranslate 18 Sep
- Grok Voice Transcribe 2.0 17 Sep
- Gemini 3.8 Live 15 Sep
- Qwen-Image-2.1 20 Sep
- Enhance Frame Rate 17 Sep
Tool releases · 56
All tools- Grok Bot 0.57.0: Bots can start Projects, cloud agents that plan and run their own agents 18 Sep
- Kimi Code 2.0.0: Desktop app launcher and mermaid diagrams in the terminal 17 Sep
- Claude: Cowork in every chat, plus designs, decks and docs anywhere 16 Sep
- Grok Build 1.0.40: Grok 4.7 model now available 20 Sep
- Claude Code 2.1.278: Auto mode now defaults to the server-side classifier at no extra cost 19 Sep
- Mistral Vibe 2.25.5: Unified Harness leaves experimental; tighter shell approvals and hooks 18 Sep
- Kilocode 7.7.4: Sessions now link to GitLab and Bitbucket pull requests too 18 Sep
- Codex 0.155.0: Experimental voice conversations and live reasoning summaries arrive in the TUI 17 Sep
- Copilot 1.0.86: Custom agents can now opt into repository instruction files 17 Sep
- ChatGPT: Connect multiple accounts to plugins, not just Google ones 17 Sep
- T3 Code 0.0.42: Rewind chats keeping file changes, multiple linked PRs, device simulators 16 Sep
56 releases across 13 tools; the ones with something new are above. Busiest: Grok Build, 9 releases.
Top stories · 24
Archive- Anthropic embeds Accenture evaluators, $1B each 18 Sep
- Claude speeds up 30+ open-source biology models 17 Sep
- OpenAI launches Astra for Law 17 Sep
- Anthropic says Claude leads 26% of its R&D 17 Sep
- OpenAI sets standing misalignment disclosure framework 16 Sep
- Gemini 3.8 Live tops speech-to-speech index 15 Sep
- xAI ships Grok Voice Transcribe 2.0 18 Sep
- Anthropic opens life sciences verification program 17 Sep
- OpenAI tests Sponsored Agents in ChatGPT 16 Sep
- OpenAI finds cross-occupation AI use recurring 16 Sep