AIME: leaderboard
National math exam given to top high-school students.
Metric: Accuracy (%). Source: www.vals.ai. Status: saturated. 96 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (High) | 98.12 |
| 2 | GPT-5.2 (xHigh) | 96.88 |
| 3 | Muse Spark | 96.88 |
| 4 | Gemini 3 Pro (Preview) (High) | 96.68 |
| 5 | GPT-5.4 (xHigh) | 96.67 |
| 6 | Grok 4.20 0309 (Reasoning) | 96.46 |
| 7 | Claude Opus 4.7 | 96.25 |
| 8 | Kimi K2.5 (Thinking) | 95.62 |
| 9 | Claude Opus 4.6 (Thinking) | 95.62 |
| 10 | GPT-5.4 Mini (xHigh) | 95.62 |
| 11 | Gemini 3 Flash (Preview) (High) | 95.62 |
| 12 | Claude Opus 4.5 (20251101) (Thinking) | 95.42 |
| 13 | Qwen 3.6 Plus | 94.58 |
| 14 | GPT-5 (High) | 93.37 |
| 15 | GLM-4.7 | 93.33 |
Interactive version: theaggregate.ai/benchmark?slug=aime · How It Works · Data refreshed daily, snapshot 2026-09-05.