U-MATH — leaderboard

Metric: Accuracy (%). Source: huggingface.co. 28 models tracked.

Top models

#ModelScore
1O186.82
2Gemini 2.0 Flash (01-21) (Thinking)83.64
3O3 Mini82.18
4DeepSeek R180.73
5O1 Mini76.27
6QwQ 32B-Preview73.09
7QVQ-72B-Preview65
8DeepSeek V362.64
9Gemini 1.5 Pro60.09
10Qwen2.5-Math-72B-Instruct59.45
11Gemini 1.5 Flash57.82
12Athene-V2-Chat54.91
13Phi-454.45
14Qwen 2.5 32B Instruct52.36
15Qwen 2.5 72B Instruct51.18

Interactive version: aibenchmarks.dev/benchmark?slug=u-math · How the rankings work · Data refreshed daily, snapshot 2026-07-08.