Mathematical Proof on IMO-ProofBench
91.9Advanced ScoreAletheia
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Aletheia2026.03 | 91.9 | — | — | |
| Gemini 3 Deep Think2026.03 | 76.7 | — | — | |
| Gemini Deep Think (IMO Gold)2026.03 | 65.7 | 89 | 76.7 | |
| DeepSeek-Math-V2-671B-A37B2026.03 | 61.9 | 99 | 80.2 | |
| DeepSeek-Math-V2-671B-A37BReproduced score=true, Judge model=DeepSeek-V3.2-Speciale, LLM prompt=LLM ProofAutoGrader prompt2026.03 | 57.7 | 99.5 | 78.6 | |
| Nemotron-Cascade-2-30B-A3BJudge model=DeepSeek-V3.2-Speciale, LLM prompt=LLM ProofAutoGrader prompt2026.03 | 53.4 | 92.5 | 72.9 | |
| GPT-5.2-Thinking (high)2026.03 | 35.7 | — | — | |
| Gemini 3 Pro2026.03 | 30 | — | — | |
| GPT-5 Pro2026.03 | 28.6 | — | — |