An unreleased OpenAI model solved more than 100 open math problems
This story was updated after publication (Oct. 9, 2026). Details
The company says the model began training on August 28 and, in under a month, solved over 100 open problems spanning most areas of mathematics. An independent advisory group of mathematicians was then convened at the Institute for Advanced Study in Princeton, including Timothy Gowers, Martin Hairer, Edward Witten and Ravi Vakil. The claim has not been independently verified: OpenAI published neither the list of problems nor the proofs. Earlier, 25 Fields medalists signed an open letter criticizing how the Navier-Stokes announcement was handled.
Sources
- OpenAI, “Advisory Group on Mathematics and Artificial Intelligence”, (openai.com)
- Terence Tao (What's new blogu), “Announcing the Advisory Group on Mathematics and Artificial Intelligence”, (terrytao.wordpress.com)
- Terence Tao (What's new blogu), “A Severe Misalignment of AI in Mathematics”, (terrytao.wordpress.com)
- OpenAI, “On the Navier–Stokes Millennium Prize Problem”, (openai.com)
Corrections and updates
- After this story was published, OpenAI released 722 manuscripts produced by the internal model on GitHub on October 6, many of them with Lean formalizations; independent verification by the mathematics community is ongoing.
About this story
This story was posted on Instagram by @jarrus.tech on Sept. 25, 2026.
Spotted an error in this story? [email protected] · Instagram
Short link: thejarrus.com/en/openai-open-problems
This story in Turkish: OpenAI’ın yayımlanmamış modeli 100’den fazla açık matematik problemi çözdü
On the same topic
Robot tests: even the best model completed only 19% of 84 robot tasks, and 63 tasks were solved by no model
In RobotWorld, a benchmark released October 7, the top model, GPT-6 Astra, completed only 16 of 84 simulated robot tasks; no model solved 63 of the tasks.
OpenAI’s image model posts a near-perfect score on text-dense images
On UltraText Bench, a dense-text test led by Westlake University, OpenAI’s GPT Image 2 ranked first of 24 configurations with 99.35 out of 100.
AWS adds Chinese lab Z.ai’s GLM 5.3 model to Amazon Bedrock
AWS added GLM 5.3, a 753-billion-parameter model from Beijing-based Z.ai (formerly Zhipu AI), to Amazon Bedrock for eligible enterprise customers on October 5.