FrontierMath Erd\H{o}s
- What changed
- FrontierMath Erdos benchmark released, evaluating five AIs on 68 open problems with a $300 budget.
- Why you should care
- Provides a standardized, fixed set of open problems for systematic AI evaluation.
- Your move
- Watch. Track future model performance on this fixed benchmark.
- Evidence
- Confirmedarxiv.org ↗
- What to watch next
- New model releases or updated evaluations that achieve non-zero scores on the FrontierMath Erdos benchmark.
- Event
- research
- Event date
- Sep 23, 2026
- Relevant to
- General AI readers