Sources and Lean proof links for the IMO 2026 model benchmark

deedydas · x · 2026-07-21

This reply to the original IMO 2026 benchmark post adds the source links for the Lean proof and the model runs.

It reinforces the same result: Claude Fable 5, GPT-5.6 Sol, Kimi K3, and Axiom all solved the contest, with Axiom providing a formalized proof in Lean. The linked screenshots also show the safety warning encountered during the run, which is part of the original benchmark context.

Related event: Multiple Frontier AI Models Achieve Perfect Scores in IMO 2026 Testing(6 posts)→

Original post →

More from Models

Models channel →