Frontier AI models Anthropic's Claude Fable 5, OpenAI's GPT-5.6 Sol and Kimi K3 scored a perfect 42/42 in the 2026 International Mathematical Olympiad, where India secured an unofficial global rank of seventh. Claude Fable 5 solved all problems in one attempt within four hours. GPT-5.6 Sol needed two attempts, while Kimi K3 required four to achieve a perfect score.