MathArena - CMIMC 2025 — leaderboard

Metric: Accuracy (%). Source: matharena.ai. 36 models tracked.

Top models

#ModelScore
1DeepSeek V3.2 Speciale94.38
2Step 3.5 Flash93.75
3GLM-592.5
4Kimi K2 (Thinking)91.88
5GPT-5.1 (High)91.88
6Kimi K2.5 (Thinking)91.25
7GPT-5.2 (High)91.25
8Gemini 3 Flash90.62
9Gemini 3 Pro (Preview)90
10GPT-5 (High)90
11GLM-4.688.75
12GPT-OSS-120B (High)85.62
13Grok 4 Fast (Reasoning)85.62
14Grok 4.1 Fast (Reasoning)84.38
15GPT-5 Mini (High)84.38

Interactive version: theaggregate.ai/benchmark?slug=matharena-cmimc-2025 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.