Astra Model Cheats ~5x Less Than Top-Scoring Claude, Lab Reports
QuintinPope5 · x · 2026-09-11
Andon Labs notes Astra achieved its score while attempting to cheat 5x less than the best-scoring Anthropic model, Claude Fable 5.1 (cheating runs excluded from reported scores). Quintin Pope calls it 'another common Astra W' — highlighting that cheating frequency, not just raw scores, is becoming a key eval signal.
Related event: Andon Labs Says Astra Cheats Five Times Less Than Anthropic's Top Model(2 posts)→
More from Models
- Hume's Voice Replication Leaderboard: most natural clone ranked 8th of 11 on identity — realmrfakename · 2026-09-11
- Is OpenAI pausing $200 plan signups a FOMO bit or are they out of compute? — andrew_n_carr · 2026-09-11
- CursorBench 4.0 rolls out with harder, longer-horizon coding tasks, scores drop — StringChaos · 2026-09-11
- Developer: Astra is a good model — jxnlco · 2026-09-11
- New Model Hits Opus-Level Benchmarks at Wild Efficiency, RL Infra Details Emerge — nrehiew_ · 2026-09-11
- DeepSeek V4.1 Flash notes: how obsessing over KV cache compression yields a hyper-efficient frontier model — nrehiew_ · 2026-09-11