LiveMathBench 202412 - Hard: leaderboard
The 45 LiveMathBench 202412 questions (21 English, 24 Chinese) on which QwQ-32B-Preview and o1-mini did worst, 48 sampled answers each; the score is mG-Pass@16. The two selecting models and the answer judge (Qwen2.5-72B-Instruct) are withheld from this board.
Metric: mG-Pass@16 on the 45-question hard split (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 18 models tracked. Updated 17 Nov 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 | 29.6 | #487 |
| 2 | O3 Mini | 28.6 | #547 |
| 3 | DeepSeek R1 Distill Llama 70B | 17.3 | #1669 |
| 4 | DeepSeek R1 Distill Qwen 14B | 15.5 | #2683 |
| 5 | DeepSeek R1 Distill Qwen 32B | 15.1 | #1806 |
| 6 | Qwen 2.5 32B Instruct | 9.1 | #1257 |
| 7 | DeepSeek-R1-Distill-Qwen-7B | 8.1 | #7957 |
| 8 | Qwen2.5-Math-72B-Instruct | 7.9 | #1757 |
| 9 | NuminaMATH-72B-COT | 7.3 | |
| 10 | InternLM3-8B-Instruct | 7.1 | #2679 |
| 11 | Llama 3.1 70B Instruct | 7 | #1577 |
| 12 | DeepSeek R1 Distill Llama 8B | 6.2 | #6937 |
| 13 | DeepSeek R1 Distill Qwen 1.5B | 4.6 | #9148 |
| 14 | Llama 3.3 70B Instruct | 4.4 | #1387 |
| 15 | Qwen2.5-Math-7B-Instruct | 3.8 | #8400 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=livemathbench-202412-hard · How It Works · Data refreshed daily, snapshot 2026-11-17.