Anthropic Opus 5 Achieves 42/42 on 2026 Math Olympiad Tasks

Anthropic’s Opus 5 reportedly scored 42/42 on 2026 International Mathematical Olympiad tasks and beat its predecessor on several major benchmarks.

It also showed strong performance on professional workloads in areas like finance, law, and medicine, including ARC AGI 3.

The model is said to cost about half as much as Claude Fable 5, making it a more cost-efficient option.

However, the article notes weaker high-level reasoning in some cases and a fallback mechanism to Opus 4.8, highlighting deployment risks.