The week in AI

1. PixVerse R2 PixVerse · Video · Big 22 Sep
2. MiMo-V2.6-Pro Xiaomi · LLM · Big 22 Sep
3. MiMo-V2.6-Flash Xiaomi · LLM · Big 22 Sep
4. Claude Opus 5.5 Anthropic · LLM · Big 22 Sep
5. Grok 4.7 xAI · LLM · Big 21 Sep
6. Gemini 3.8 Flash TTS Google · Audio · Medium 22 Sep
7. GPT-6 Sol OpenAI · LLM · Medium 22 Sep
8. Recraft V4.1 Flash Recraft · Image · Light 23 Sep
9. Gemini 3.8 Flash-Lite TTS Google · Audio · Light 22 Sep
10. GPT-6 Luna OpenAI · LLM · Light 22 Sep

1. Grok Bot 0.61.0: Team Bots let you share one Bot with your whole team and Slack 26 Sep
2. Cursor: Two new bots: Rollouts watches deploys, Security Review checks PRs 23 Sep
3. Grok Build 1.0.43: MCP tool prompts fixed in minimal mode, plus unanswered-prompt alerts 27 Sep
4. Claude Code 2.1.283: New admin controls to allow or block models, plus a prompt audit 25 Sep
5. Kilocode 7.8.0: Agent Manager browser previews can now load public HTTPS pages 25 Sep
6. Codex 0.157.0: GPT-6 Sol and Luna models arrive, with fullscreen transcripts on by default 25 Sep
7. Claude: Submit plugins to the Claude directory via a new developer portal 25 Sep
8. ChatGPT: Security history shows recent sign-ins and security changes 25 Sep
9. DeepSeek Harness 0.1.7-rc.2: Scheduled tasks, desktop onboarding and customizable keyboard shortcuts 24 Sep
10. Kimi Code 2.1.1: Filesystem watchers back on by default; 2.1.0 defensive changes rolled back 24 Sep
11. Mistral Vibe 2.25.8: Archive local sessions, plus many Rust CLI and security fixes 23 Sep
12. Copilot 1.0.88: Fork mid-turn, terminal notifications and many MCP fixes 22 Sep

46 releases across 14 tools; the ones with something new are above. Busiest: OpenCode, 7 releases.

1. 1 Anthropic releases Claude Opus 5.5 Models 26 Sep
2. 2 Anthropic launches Claude Fable 5.1 and Mythos 5.1 Models 26 Sep
3. 3 Claude finds novel CRISPR-like enzyme system World 23 Sep
4. 4 OpenAI forms independent mathematics advisory group World 21 Sep
5. 5 OpenAI launches cheaper GPT‑6 Sol and Luna Models 22 Sep
6. 6 Xiaomi open-sources MiMo-V2.6, an RL-scaled model Models 22 Sep
7. 7 xAI Ships Grok 4.7 at Unchanged Prices Models 21 Sep
8. 8 Claude computes nine-loop N=4 super-Yang-Mills amplitude World 25 Sep
9. 9 Altman pledges AI restraint at UN Security Council World 23 Sep
10. 10 OpenAI opens MentalHealthBench, built with 80+ clinicians Benchmarks 23 Sep

1. 45 Week 45 · 2–8 Nov 2026 8 models · 36 releases · 48 stories
2. 44 Week 44 · 26 Oct – 1 Nov 2026 7 models · 57 releases · 48 stories
3. 43 Week 43 · 19–25 Oct 2026 9 models · 47 releases · 33 stories
4. 42 Week 42 · 12–18 Oct 2026 7 models · 45 releases · 42 stories
5. 41 Week 41 · 5–11 Oct 2026 7 models · 36 releases · 32 stories
6. 40 Week 40 · 28 Sep – 4 Oct 2026 12 models · 62 releases · 43 stories
7. 39 Week 39 · 21–27 Sep 2026 10 models · 46 releases · 36 stories
8. 38 Week 38 · 14–20 Sep 2026 8 models · 56 releases · 24 stories