Open LMM Reasoning - DynaMath - Subject-puzzle test: leaderboard
The "DynaMath - Subject-puzzle test" subtask of Open LMM Reasoning. A task from the Open LMM Reasoning leaderboard, which evaluates large multimodal models on multi-step visual, mathematical and logical reasoning problems.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 100 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 Mini | 47.1 | #355 |
| 2 | GPT-4.1 (2025-04-14) | 41.2 | #250 |
| 3 | Claude 3.5 Sonnet (20241022) | 35.3 | #291 |
| 4 | InternVL3-38B | 35.3 | #405 |
| 5 | Doubao-1.5-Pro | 35.3 | #287 |
| 6 | Gemini 2.0 Flash | 29.4 | #342 |
| 7 | InternVL3-78B | 29.4 | #352 |
| 8 | Claude 3.7 Sonnet | 29.4 | #249 |
| 9 | GPT-4o ChatGPT | 29.4 | |
| 10 | Gemma 3 12B | 23.5 | #671 |
| 11 | InternVL2.5-78B-CoT | 23.5 | #469 |
| 12 | Grok 2 (1212) | 17.6 | #496 |
| 13 | QVQ-72B-Preview | 17.6 | #440 |
| 14 | GPT-4.1 Nano | 11.8 | #731 |
| 15 | InternVL3-8B | 11.8 | #620 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-lmm-reasoning-dynamath-subject-puzzle-test · How It Works · Data refreshed daily, snapshot 2026-10-20.