DeathMath - Math: leaderboard
DeathMath (DOoM leaderboard) by Vikhr Models, math section: complex mathematics problems posed in Russian, answered with a Russian system prompt. Accuracy 0-100, higher is better; one published run per model, with system prompts that differ between submissions.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 32 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 94 | #82 |
| 2 | GPT-5 | 91 | #105 |
| 3 | O3 Mini (High) | 88.4 | #282 (O3 Mini) |
| 4 | Gemini 2.5 Pro | 87.4 | #159 |
| 5 | O3 | 86.8 | #132 |
| 6 | O4 Mini (High) | 86.8 | #185 (O4 Mini) |
| 7 | GPT-OSS-120B | 84.9 | #346 |
| 8 | GPT-5 Mini | 84.9 | #193 |
| 9 | GPT-5 Nano | 83.9 | #439 |
| 10 | GPT-OSS-20B | 78.9 | #523 |
| 11 | Claude Opus 4.1 | 70.4 | #146 |
| 12 | Claude Sonnet 4.5 | 68.8 | #148 |
| 13 | QwQ-32B | 65.3 | #422 |
| 14 | Claude Sonnet 4 | 63.3 | #207 |
| 15 | GPT-4.1 | 58.4 | #258 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=deathmath-math · How It Works · Data refreshed daily, snapshot 2026-11-05.