aiaimodel.cards
← All benchmarks

AIME 2025 (with tools)

Problems from the American Invitational Mathematics Examination, a hard high-school math competition.

Your workloadRAG over docs · 12K in · 600 out · 50K requests/mo · 30% cached
  1. 1gpt-oss-20b · OpenAIbest value
    98.7%
  2. 2gpt-oss-120b · OpenAI
    97.9%

“Best value”: no other model here reports a higher score at a lower cost for your workload. Scores are self-reported by each maker with different settings, so treat close calls as ties.

How each maker reported it

  • gpt-oss-20b Figures 1, 2, 4; Table 2

    OpenAI's own pass@1 results at reasoning level high. AIME and HLE with tools; Codeforces with a terminal tool. o3-mini was run without tools on AIME, HLE and Codeforces, so those cells are left empty. Baseline values read from OpenAI's charts.

  • gpt-oss-120b Figures 1, 2, 4; Table 2

    OpenAI's own pass@1 results at reasoning level high. AIME and HLE with tools; Codeforces with a terminal tool. o4-mini and o3 values read from OpenAI's charts.

Only each model's own maker's number is shown; scores a maker reports for rival models are left out.