Breaking

OpenAI Says Internal Model Solved 100-Plus Math Problems as Independent Advisers Weigh Release

October 6, 2026 at 13:32 EDT

Foundation Models
AI Agents
Research & Papers

OpenAI says a newly trained internal model has resolved more than 100 long-standing open problems in mathematics, intensifying a debate over how AI-generated proofs should be verified, credited, and released. The company has enlisted an independent Advisory Group on Mathematics and Artificial Intelligence as it prepares to coordinate the disclosure of a large body of reported results.

September 2026 · OpenAI

100-plus math results await release

OpenAI says an internal model resolved more than 100 long-standing open problems, turning attention to verification, attribution, and how AI-generated mathematics should enter the field.

100+

reported open problems

initial advisers

600+

survey responses

17h

Lean verification

Agent scale in the proof push

~100Euler agents

~10,000Navier-Stokes agents

The taller column is scaled to the reported concurrent-agent order of magnitude.

Train

A new internal model starts training on August 28.

Generate

Agent groups produce proofs and formal artifacts.

Release

Independent advisers weigh how results should be shared.

Practical upside

Researchers describe AI tools as efficient assistants for writing, checking ideas, and exploring proof paths.

Core concern

Mathematicians warn that verification, attribution, access, and human understanding cannot be skipped.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.