Skip to content

ArXivMath August 2026 with tools

The August 2026 ArXivMath release evaluated with a code-execution sandbox and no internet access.

2 models

Higher is better.

  1. 1Claude Opus 5.5Anthropic96.9
  2. 2Claude Sonnet 5.5Anthropic95.2
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Final-answer accuracy with tools
Tasks
57 recent research-mathematics problems
Authors
MathArena and Anthropic (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.