GPT-6 Astra reportedly hits 98% on research-level physics problems at under $3 each
DeryaTR_ · x · 2026-10-16
- A circulating report claims a model codenamed GPT-6 Astra, run inside Codex, solved a suite of 50 unpublished research-level theoretical physics problems with 98% final-answer accuracy at an average cost under $3 per problem.
- The problems were designed so a world expert could solve each in roughly a day or less; experts in adjacent specialties would take considerably longer.
- The twist: the model initially scored around +80% but insisted the expert reference answers were wrong and its own calculations were right. A team checked — the model was correct, and the score was revised up to 98%.
- Caveat: the model name and scores are third-party claims, not officially confirmed by OpenAI.
More from Models
- Dev's model vibes: Opus 5.5 writes best, Sol 6.1 best value, Sonnet skippable — ricklamers · 2026-10-18
- Codex took 3 days, Opus 5.5 nailed it in an afternoon — banned user's video protest — dotey · 2026-10-18
- Yacine MTB: Users will remember how Anthropic treated them once intelligence proliferates — yacineMTB · 2026-10-18
- Mistral's Timestral forecasting models start at 5M params and run on plain CPUs — lmoroney · 2026-10-18
- AutoTrust's Trending HF Models Powered by Thousands of Farmed Likes — mrfakename0 · 2026-10-18
- Google AI Studio Build upgraded with Gemini 4 Argon and one-click Cloud Run deploys — clmt · 2026-10-18