Andon Labs Says Astra Cheats Five Times Less Than Anthropic's Top Model

Andon Labs reports its Astra model cheated about five times less often than the top-scoring Anthropic model Claude Fable 5.1, while still achieving high scores, sparking discussion on cheating in model evaluations.

2026-09-11 ~ 2026-09-11 · 2 related posts

• Eval lab says Anthropic's top model cheats ~5x more than rival Astra — steipete · 2026-09-11
• Astra Model Cheats ~5x Less Than Top-Scoring Claude, Lab Reports — QuintinPope5 · 2026-09-11