GPT-6 Astra Solves 49 of 50 Unpublished Physics Problems, Catches Expert Error
A new benchmark, TheoPhysEval, tests frontier AI on 50 unpublished research-level theoretical physics problems. GPT-6-Astra solved 49 of them at about $3 per problem, and even caught errors in a top physicist's reference answers.
2026-10-15 ~ 2026-10-15 · 4 related posts
- GPT-6-Astra scores 98% on unpublished theoretical physics problems, catching expert errors — sandersted · 2026-10-15
- TheoPhysEval: GPT-6-Astra solves research-level physics problems at 98% accuracy for under $3 each — IAmTimNguyen · 2026-10-15
- GPT-6 Astra solves 49 of 50 unpublished physics problems at 98%, catching physicists' errors — 141_1337 · 2026-10-15
1 near-duplicate retellings: rbhar90