Epoch AI benchmark: no LLM hits human employee standard, Fable 5.2 tops at 82%
Jsevillamol · x · 2026-11-18
Epoch AI tested recent frontier models on Epoch Automation Reports, a benchmark of open-ended real work tasks (Epoch-style diagrams, basic research experiment design), scored by Epoch employees against internal quality standards.
- No model met the Epoch employee standard (100%)
- Fable 5.2 scored highest, averaging 82%
- GPT-6.2 Astra ranked fourth at 65% despite sharing the top ECI score of 174 with Fable 5.2
Kelly Hong adds that despite strong benchmark performance, all models show weak judgment and presentation, reusing existing ideas and failing to iterate even on simple work-trial research projects.
More from Models
- Pet theory: Anthropic is skipping looping agents for safety and monitorability — NathanpmYoung · 2026-11-18
- SayGM lists GPT-6.2 Astra 4 hours after launch, 12.4% under OpenAI list price — markjeffrey · 2026-11-18
- Early tester: no negatives so far in GPT 6.2 Astra and Sol hands-on — craigsdennis · 2026-11-18
- SpaceX CFO praises Grok Bot as 'probably the best agentic AI solution out there' — elonmusk · 2026-11-18
- Claude makes art and reflects on itself in open-ended creativity experiment — andybdesign · 2026-11-18
- Microsoft AI unveils seven cost-effective models and second-gen Maia 200 accelerator — iJeff · 2026-11-18