AIME 2025 (Artificial Analysis avg@10)
Math · 2025-02-12
Artificial Analysis' AIME 2025 protocol averages pass@1 over ten independent attempts for each of 30 questions (300 scored attempts). It is separated from single-attempt AIME observations so its uncertainty is not modeled as a 30-item binomial proportion.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.2 | 99.0 |
| GPT-5-Codex | 98.7 |
| Gemini 3 Flash Preview | 97.0 |
| DeepSeek-V3.2-Speciale | 96.7 |
| MiMo-V2-Flash | 96.3 |
| Gemini 3 Pro Preview | 95.7 |
| GPT-5.1-Codex | 95.7 |
| GLM-4.7 | 95.0 |
| KAT-Coder-Pro V1 | 94.7 |
| Kimi K2 Thinking | 94.7 |
| GPT-5 | 94.3 |
| Nova 2 Lite | 94.3 |
| GPT-5.1 | 94.0 |
| gpt-oss-120b | 93.4 |
| Grok 4 | 92.7 |
| DeepSeek-V3.2 | 92.0 |
| GPT-5.1-Codex mini | 91.7 |
| Opus 4.5 | 91.3 |
| Nemotron 3 Nano 30B A3B | 91.0 |
| Qwen3 235B A22B Thinking 2507 | 91.0 |
| GPT-5 Mini | 90.7 |
| O4 Mini | 90.7 |
| K-EXAONE 236B-A23B | 90.3 |
| DeepSeek-V3.1 | 89.7 |
| DeepSeek-V3.1-Terminus | 89.7 |
| Grok 4 Fast | 89.7 |
| Nova 2 Omni (Preview) | 89.7 |
| gpt-oss-20b | 89.3 |
| Grok 4.1 Fast | 89.3 |
| Ring-1T | 89.3 |
| Nova 2 Pro (Preview) | 89.0 |
| o3 | 88.3 |
| Qwen3 VL 235B A22B Thinking | 88.3 |
| Apriel-1.6-15B-Thinker | 88.0 |
| Sonnet 4.5 | 88.0 |
| INTELLECT-3 | 88.0 |
| DeepSeek-V3.2-Exp | 87.7 |
| Gemini 2.5 Pro | 87.7 |
| Apriel-1.5-15B-Thinker | 87.5 |
| GLM 4.6 | 86.0 |