OpenAI opens AI-generated mathematics results for review
The current GitHub collection contains 719 manuscripts across 372 result families. Verification levels vary, and about 42% of top-line results have Lean formalizations.
AT A GLANCE
Key points
- 719 manuscripts across 372 result families
- About 42% of top-line results have Lean formalizations
- The model was tested on approximately 4,000 problems
- An average result used about three hours of ChatGPT Pro thinking compute
What is changing?
OpenAI released results generated by an unreleased internal frontier model on open mathematics problems through GitHub on October 6. The current repository contains 719 manuscripts across 372 result families, grouping principal results with companion arguments, consequences, and alternative proofs.
OpenAI explicitly says the papers are at different stages of verification and that unformalized results may contain issues. According to the repository, about 42% of top-line results have Lean formalizations, while corrections and version changes remain in a public history.
The model was evaluated on approximately 4,000 problems. The company says an average result used the equivalent of roughly three hours of ChatGPT Pro thinking compute and also released ten abridged reasoning summaries plus attempt statistics.
The release does not mean every claim has been validated, and it does not make the model generally available. Its main significance is opening large-scale AI research output to community review with versioning, citation, and formal-verification mechanisms.
Explore the original sources
Source date: 6 October 2026.
Published by AI Radar: 7 Oct 2026, 19:21 (Istanbul, UTC+3). Updated: 9 Oct 2026, 23:11.
FOLLOW THE TOPIC
Topics and tags
Share your discovery.
Send this story to a friend.
Reader comments
What do you think about this development? Share your experience or question. Comments are published after editorial approval.
Your name/nickname and comment become public if approved. Read: Privacy notice.
Loading comments…