Astra Crushes Opus 15-4 in Chess Despite 360-Point Lower Elo

Peter Gostev's match data shows Google's Astra leading Claude Opus 15-4 in chess with a five-game win streak, even though Opus's Elo rating is about 360 points higher after both non-reasoning models learned from 200 games against Stockfish.

2026-10-08 ~ 2026-10-10 · 2 related posts