Checked Oct 8, 2026Start a new check

What was checked

Suddenly, AI just solved 90 of the 500 most important open problems in mathematics

Assessment summary

Needs caution

The figure of 90 refers to an outside estimate involving roughly 500 selected problems, not an authoritative count of fully solved problems. The reported results include partial advances, and the claims have not yet received broad independent confirmation.

Key findings

Model assessments

Assessment summary

The statement appears to summarize an outside tracker’s interpretation of OpenAI’s newly released mathematical results, not a confirmed scientific consensus.techmeme.com startupfortune.com Reports describe full and partial results across many problems, but the claimed total of 90 and the word “solved” remain insufficiently verified, with no broad independent mathematical confirmation.gigazine.net startupfortune.com

Claims reviewed

AI solved 90 of the 500 most important open problems in mathematics.

Needs context
The figure of 90 refers to an outside estimate involving roughly 500 selected problems, not an authoritative count of fully solved problems.techmeme.com startupfortune.com The reported results include partial advances, and the claims have not yet received broad independent confirmation.techmeme.com gigazine.net

This claim may mislead readers without added context.

Assessment summary

  • OpenAI did publish 722 manuscripts on October 6, claiming to cover 90 of around 500 problems from mathematics' hardest open-problem lists.
  • However, the claim uses "solved" without important caveats.
  • The shift from "we made progress on problems" to "we cleared 90 of the field's hardest 500" is the headline, and it's the number readers are now trying to check.
  • The word "suddenly" also overstates the timeline, as this represents months of incremental progress.

Assessment summary

  • The headline-like statement is misleading.
  • Recent online posts and a company release claim models produced solutions or progress on roughly 90 of a 500-item ranked list, but independent verification, peer review, and accepted proofs are lacking for most entries.
  • Many items appear to be partial matches, formalizations, or model-generated attempts tied to a specific curated list rather than widely recognized, formally accepted solutions.

Claims reviewed

AI just solved 90 of the 500 most important open problems in mathematics

Needs context
Public reports and a dataset/repository circulated claiming ~90 matches against a 500-problem list, but those matches are not equivalent to universally accepted, peer-reviewed solutions; many are partial, formalization artifacts, or depend on a specific LLM-ranked list rather than community consensus.

This claim may mislead readers without added context.

Sources cited

26 sources

Share this result