Atlas

Benchmarks

← All benchmarks

ArXivMath 05/2026

Math · 2026-06-11

This MathArena track evaluates models on the ArXivMath 05/2026 problem set. Scores report the percentage of problems answered correctly.

Top models (higher is better)

ModelScore
Claude Fable 587.5
GPT-5.577.5
Opus 4.865.0
Kimi K361.7
GLM-5.256.7
DeepSeek-V4-Flash55.8
Gemini 3.6 Flash54.2
GLM-5.152.5
Gemini 3.1 Pro Preview51.7
Step 3.7 Flash46.7
Gemini 3.5 Flash44.2
Qwen3.6 35B A3B37.5
Qwen3.5-2B10.6
Loading Atlas data…