Skip to content

PolyMath

A multilingual mathematical reasoning benchmark that tests whether math performance transfers across languages rather than only in English.

6 models

Higher is better.

  1. 1Qwen3.7 MaxAlibaba86.5
  2. 2Qwen3.7 PlusAlibaba84.0
  3. 3Claude Opus 4.5Anthropic79.0
  4. 4Qwen3.5 397BAlibaba73.3
  5. 5K-EXAONE 2.0LG AI Research71.3
  6. 6Kimi K2.5Moonshot AI43.1
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Cross-lingual mathematical reasoning
Tasks
Multilingual math problems
Authors
Qwen (2026)

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