U-MATH - Sequences & Series: leaderboard
The Sequences & Series subtask of U-MATH (Toloka) — university-level sequences-and-series problems; accuracy (%).
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 34 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 Mini | 93.51 | #272 |
| 2 | DeepSeek R1 | 92.86 | #251 |
| 3 | O1 | 92.86 | #183 |
| 4 | Gemini 2.0 Flash (01-21) (Thinking) | 88.31 | #322 (Gemini 2.0 Flash (01-21)) |
| 5 | QwQ 32B-Preview | 84.42 | #651 |
| 6 | O1 Mini | 76.62 | #323 |
| 7 | DeepSeek V3 | 72.08 | #321 |
| 8 | Qwen2.5-Math-72B-Instruct | 68.83 | #717 |
| 9 | QVQ-72B-Preview | 66.23 | #440 |
| 10 | Gemini 1.5 Pro | 62.99 | #445 |
| 11 | Gemini 1.5 Flash | 60.39 | #618 |
| 12 | Qwen 2.5 32B Instruct | 56.49 | #498 |
| 13 | Mistral Large 2 (Nov) Instruct (2411) | 55.19 | #465 |
| 14 | Qwen 2.5 72B Instruct | 54.55 | #454 |
| 15 | Qwen2.5-Math-7B-Instruct | 51.95 | #1444 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=u-math-sequences-series · How It Works · Data refreshed daily, snapshot 2026-10-20.