OpenAI's 721 AI math papers: most results not yet checked
Why it matters
Evidence on whether AI can do new research math. OpenAI says an internal model produced the results, and computer checkers had confirmed 300 of 719 by 7 Oct. It says unchecked ones could have mistakes.
The facts
- Confirmed
OpenAI posted 722 math papers on GitHub, a public code site. It says an internal AI model produced the results.
- Confirmed
One paper claims progress on the Riemann hypothesis, a famous unsolved problem. The paper itself says it does not solve it.
The Quasi-Riemann Hypothesis (OpenAI, 30 Sep 2026), Section 1
Behind the headline
- Confirmed
OpenAI says computer proof checkers have confirmed 300 of its 719 main results (about 42%), and the unchecked ones could have mistakes.
- Confirmed
OpenAI withdrew 3 papers after a sign error (a wrong plus or minus), leaving 719, and corrected 14 others.
How it unfolded
- 30 Sep 2026OpenAI dates its Riemann hypothesis paper. Source: GitHub
- 6 OctOpenAI posts 722 papers on GitHub. Source: GitHub
- 7 OctOpenAI withdraws 3 papers over a sign error and fixes 14. It says computer checkers have confirmed 300 of 719 results. Source: GitHub