Atlas

Benchmarks

← All benchmarks

ArXivMath 03/2026

Math · 2026-04-01

This MathArena track evaluates models on the ArXivMath 03/2026 problem set. Scores report the percentage of problems answered correctly.

Top models (higher is better)

ModelScore
GPT-5.577.5
Opus 4.875.0
Gemini 3.1 Pro Preview68.3
GPT-5.467.8
Opus 4.662.5
GLM-5.261.7
DeepSeek-V4-Pro55.8
Gemini 3.5 Flash55.8
Kimi K2.655.8
DeepSeek-V4-Flash52.5
Opus 4.750.8
GLM-5.149.2
GLM-545.8
Step 3.7 Flash45.0
Step 3.5 Flash43.3
Qwen3.5-2B9.8
Loading Atlas data…