Leak: Anthropic's Internal Model 2 Outperforms, RSI Accelerating
teortaxesTex · x · 2026-08-16
A leak claims Anthropic's RSI (Recursive Self-Improvement) is emerging, with internal Model 2 scoring 62.8% on CoBench v2 (real AI research tasks), 12.5% higher than current Mythos 5, and already used at scale internally. The claim is unverified but sparks discussion on AI progress speed.
Related event: Rumor: Anthropic Uses Secret 'Model 2' Stronger Than Mythos 5(21 posts)→
More from Models
- Test finds Flash-0731 overfitting; DS free web app praised for speed — teortaxesTex · 2026-08-16
- Leaked specs put Kimi K3 at 2.8T params — and why a 27B dense model isn't small — Xianbao_QIAN · 2026-08-16
- 397-Billion-Parameter AI Runs on iPhone: Video Explains MoE Design and Limitations — DJTRENDSETTA · 2026-08-16
- Grok 4.6 Tops RuntimeWire Benchmark, Beating GPT-5.6 and Claude — elonmusk · 2026-08-16
- SimpleBench Results: Odd Performance from Kimi-K3, Strong Show by Grok 4.6 — scaling01 · 2026-08-16
- DeepSeek on Usability: Our Models Are Built Primarily for Ourselves — teortaxesTex · 2026-08-16