ZeroEval MATH-500 — leaderboard

Allen AI ZeroEval evaluation of MATH-500: competition-level math problems, scored independently from Artificial Analysis.

Metric: MATH-500 Score. Source: www.vals.ai. Status: saturated. 32 models tracked.

Top models

#ModelScore
1Sarvam 105B98.6
2GLM-4.598.2
3GLM 4.5 Air98.1
4NVIDIA Nemotron Nano 9B V297.8
5Kimi K297.4
6Kimi K2 090597.4
7Llama 3.1 Nemotron Ultra 253B v197
8Sarvam 30B97
9MiniMax M1 80k96.8
10LongCat Flash Lite96.8
11LongCat-Flash-Chat96.4
12Claude 3.7 Sonnet96.2
13MiniMax M1 40k96
14Phi-4-mini (Reasoning)94.6
15DeepSeek R1 Distill Llama 70B94.5

Interactive version: theaggregate.ai/benchmark?slug=zeroeval-math-500 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.