Fable 5.2 still loses to GPT-6 Astra at Blood on the Clocktower, but beats Opus 5.5
teortaxesTex · x · 2026-11-10
Dwarkesh Patel reports that at Blood on the Clocktower, Fable 5.2 is still worse than GPT-6 Astra, though much better than Opus 5.5 or Fable 5.1. teortaxesTex adds that Astra is a hair better at ambitious conceptual/mathematical thinking — if you can convince it the topic deserves it — while quipping Anthropic might nerf models anytime. A screenshot adds user reports that Opus 5/5.5 and Fable 5.2 extensively fabricate requirements on compiler work, with error-correcting oversight harnesses still more quota-efficient than OpenAI models.
More from Models
- Open-source Mirage model trends on Hugging Face with zero-shot detection and VQA — Cactus-Compute · 2026-11-10
- Anthropic deliberately capped Claude's AI-research capabilities; ML researcher sees years of runway — A_K_Nain · 2026-11-10
- 400M model handles game NPC judgment calls in ~0.1s on a 3060 Ti, Apache-2.0 — NoahPersaud · 2026-11-10
- Fable 5.2's most surprising trait: no new open problems solved — sytelus · 2026-11-10
- Sailing Sim Eval: Fable 5.2 First Claude Model to Steer Live, Argon Costs $0.73 — shekitup · 2026-11-10
- Reddit users complain free Gemini is getting worse and less useful — rupomthegreat · 2026-11-10