MathArena - SMT 2025: leaderboard
Stanford Math Tournament 2025: subject-specific tests and team rounds in algebra, calculus, geometry, and discrete math for high school students.
Metric: Accuracy (%). Source: matharena.ai. Saturation forecast: Estimated already saturated. 43 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 93.4 | #66 |
| 2 | Gemini 3 Flash | 92.92 | #99 |
| 3 | GPT-5.2 (High) | 91.98 | #110 (GPT-5.2) |
| 4 | GPT-5 (High) | 91.98 | #98 (GPT-5) |
| 5 | Step 3.5 Flash | 91.51 | #270 |
| 6 | GLM-5 | 91.04 | #141 |
| 7 | Kimi K2 (Thinking) | 91.04 | #243 (Kimi K2) |
| 8 | GPT-5.1 (High) | 91.04 | #135 (GPT-5.1) |
| 9 | GLM-4.6 | 90.57 | #252 |
| 10 | Kimi K2.5 (Thinking) | 90.57 | #144 (Kimi K2.5) |
| 11 | GPT-5 Mini (High) | 89.15 | #186 (GPT-5 Mini) |
| 12 | DeepSeek V3.2 Speciale | 89.15 | #128 |
| 13 | O4 Mini (High) | 88.68 | #180 (O4 Mini) |
| 14 | DeepSeek V3.2 (Thinking) | 87.74 | #205 (DeepSeek V3.2) |
| 15 | GPT-OSS-120B (High) | 87.74 | #341 (GPT-OSS-120B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=matharena-smt-2025 · How It Works · Data refreshed daily, snapshot 2026-10-20.