IUMB: leaderboard
16 introductory undergraduate math problems across topology, number theory, geometry, group theory, probability, real analysis, and combinatorics. Models tested with 3 attempts each (5 for avg@5 variants).
Metric: Score (%). Source: pellaml.github.io. Saturation forecast: Around January 2027. 55 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 Pro | 100 | #17 |
| 2 | GPT-5.4 Pro (xHigh) | 97.9 | #32 (GPT-5.4 Pro) |
| 3 | GPT-5.2 Pro | 89.6 | #61 |
| 4 | GPT-5.5 | 88.8 | #31 |
| 5 | Gemini 3 Deep Think | 87.5 | #27 |
| 6 | Gemini 3.1 Pro (Preview) | 85.4 | #58 |
| 7 | GPT-5.4 (Thinking) | 81.2 | #79 (GPT-5.4) |
| 8 | Claude Opus 4.6 (Thinking) | 79.2 | #63 (Claude Opus 4.6) |
| 9 | GPT-5.2 (High) | 77.1 | #110 (GPT-5.2) |
| 10 | GPT-5.2 (Thinking) | 70 | #110 (GPT-5.2) |
| 11 | GPT-5.2 (xHigh) | 68.8 | #110 (GPT-5.2) |
| 12 | Muse Spark | 66.7 | #64 |
| 13 | Gemini 3 Pro (Preview) | 62.5 | #66 |
| 14 | DeepSeek V3.2 Speciale | 62.5 | #128 |
| 15 | GLM-4.7 | 58.3 | #192 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=iumb · How It Works · Data refreshed daily, snapshot 2026-10-20.