OpenAI's GPT-6 Astra tops ErdosBench math benchmark despite no math focus
OpenAI unveiled GPT‑6 Astra, which claimed the top spot on the ErdosBench suite of 226 open mathematics problems. The model solved 106 problems, fully resolving 43 and disproving 27, earning a 3.23 score that outstripped its predecessor Sol, which solved 78. Although Fable 5.1 later reclaimed first place with fewer solves, Astra’s performance marks a clear quantitative gain in AI‑driven math…
Key points
- GPT‑6 Astra solved 106 of 226 ErdosBench math problems, fully solving 43 and disproving 27, scoring 3.23.
- OpenAI chose not to optimize Astra for math, focusing resources on recursive self‑improvement and alignment research.
- Experts caution that rapid AI‑generated proofs may overwhelm mathematicians, turning proof scarcity into overload.
Chief scientist Jakub Pachocki explained that OpenAI deliberately did not prioritize mathematics when training Astra, opting instead to allocate resources to recursive self‑improvement (RSI) and AI alignment work. The decision highlights a “spiky” development trajectory where advances in one capability come at the expense of others, fueling debate about safety and the realistic path toward AGI. Mathematicians such as Terence Tao warn that accelerating AI‑generated proofs could overwhelm the community, shifting the discipline from proof scarcity to proof overload.
Model page: GPT-6-Astra →
The story so far
3 episodes →- OpenAI's GPT-6 Astra tops ErdosBench math benchmark despite no math focusthis story
GPT-6 Astra gives mathematicians a breather, and OpenAI says that's by design
The Decoder · 10 September 2026
Loading the full article…
This text was published by The Decoder and written by Matthias Bastian. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
4sources- Inside the generational clash over AI in math: 'We don’t have the luxury of waiting'Press · businessinsider.com ·
- OpenAI vs Mathematicians: Why the AI giant is facing growing backlash — ‘threat to intellectual work’Press · livemint.com ·
- OpenAI’s feud with mathematicians is only escalatingPress · TechCrunch AI ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Liquid AI releases d1-ultra decision model at $0.042 per million tokens · 1 src
- Meta releases Muse Spark Open, an 816B-parameter open-weight multimodal model with 42B active parameters and 1M-token context · 21 src
- Google launches Gemini 4 Argon model globally for Pro and Ultra subscribers with introductory API pricing · 32 src
- Microsoft AI chief says in-house model beats OpenAI's using quarter of compute · 1 src
- NVIDIA releases Nemotron 3.5 Super VL and Bolt edge models for enterprise agents · 8 src
Comments
via GitHub Discussions