Unverified50% confidenceFactExact time
Within weeks, OpenAI's internal model successively solved multiple open problems that mathematicians had long failed to solve, leaping from barely usable to a level comparable to top human mathematicians.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/20/2026
First Seen
Valid until: 1/18/2027
Sources
AI Breaks Into Mathematics: OpenAI's Problem-Solving Sparks a Schism in Academia
youtubeThe Verge10/20/2026
Related Entities
Related Claims
VerifiedOpenAI recently disclosed its models' performance on mathematical problem solving, involving attempts at reasoning and proving complex mathematical propositions.79% similarUnverifiedOpenAI published a batch of mathematical results produced by internal frontier models, comprising 722 manuscripts; the models were posed approximately 4,000 open research questions, with each result consuming an average of about 3 hours of ChatGPT Pro reasoning compute.73% similarUnverifiedOpenAI offered three options, one of which required that when Buckmaster published his paper, he acknowledge that an internal OpenAI model had solved the problem.73% similarUnverifiedReports on Hacker News and other tech communities claim that an internal OpenAI model codenamed 'Astra' solved 10 major open problems in mathematics and computer science72% similarVerifiedOpenAI's AI systems have already solved more than 100 open problems in mathematics71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1077327API
curl https://kongchang.com/api/v1/knowledge/claims/1077327MCP
get_claim(id=1077327)