Benchgen

AIME 2025 — Results

RankModelScore
1granite-4-2-30b89.17
2granite-4-2-8b86.67
3granite-4-2-3b78.33
4lfm2-5-2-6b51.87
5k2-horizon-0-9b41.7
6gemini-3-pro1
7gpt-5-2-pro-2025-12-111
8gpt-5-21
9kimi-k2-thinking-09051
10claude-opus-4-60.998
11gemini-3-flash0.997
12seed-2-0-pro0.983
13kimi-k2-50.961
14deepseek-v3-2-speciale0.96
15glm-4-70.957
16gpt-50.946
17gpt-5-1-instant0.94
18gpt-5-1-thinking0.94
19gpt-5-10.94
20deepseek-v3-2-thinking0.931
21deepseek-v3-20.931
22seed-2-0-lite0.93
23o4-mini0.927
24gpt-oss-120b-high0.925
25qwen3-235b-a22b-thinking-25070.923

AIME 2025 — Phases

1 phaseActive

Phase 1·

AIME 2025

Aug 3, 2026
Current