Leaked chart claims Anthropic's internal Model 4 hit 178.76 AECI
haider1 · x · 2026-11-11
An unverified leak claims Anthropic's internal Model 4 scored 178.76 on the AECI benchmark, 4 points above top public model Opus 5.5 (174.6). The poster infers Mythos 5.2 is a checkpoint of internal Model 3 (AECI 177) deliberately nerfed on AI research, citing a quote that it was capped at 1.88x while a 2x-capable model stays unreleased. An attached METR-style chart shows benchmark scores accelerating at 1.9x the pre-Oct-2025 rate (95% CI 1.5x–2.8x) once internal models are included, with yearly gains rising from +15 to +29. Not confirmed by METR or Anthropic.
Related event: Rumor: Anthropic's Internal Model Leads Public Version by About 4 Points(2 posts)→
More from Models
- gnukeith details roadmap: alignment model, main model, decision model, then sub-9B specialists — gnukeith · 2026-11-11
- Indie dev gnukeith teases free self-trained LLM landing mid-January with a companion alignment model — gnukeith · 2026-11-11
- Yuanbao-Reranker: 0.6B/4B/8B Qwen3-based rerankers trained on synthetic preferences, open-sourced — _reachsumit · 2026-11-11
- Indie trainer gnukeith bets on sub-9B models: a 2B model can likely do 75% of your tasks — gnukeith · 2026-11-11
- Dev trains small CoreML decision model, beats his baselines by a wide margin — cyrilzakka · 2026-11-11
- Fable 5.2 audits an Opus 5.5-built codebase and finds issues Opus missed — CtrlAltDwayne · 2026-11-11