Open LMM Reasoning - WeMath - InsufficientKnowledge (Loose): leaderboard
The "WeMath - InsufficientKnowledge (Loose)" 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. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 Mini | 97.7 | #355 |
| 2 | Llama 3.2 11B Instruct | 80 | #1126 |
| 3 | Qwen 2 VL 2B | 65.9 | #1220 |
| 4 | InternVL2.5-2B | 63.8 | #1249 |
| 5 | Aquila-VL-2B | 54.9 | #1238 |
| 6 | Qwen 2 VL 7B | 50.1 | #750 |
| 7 | InternVL2-8B | 47 | #817 |
| 8 | InternVL2.5-8B-BoN-8 | 45.9 | #639 |
| 9 | Gemma 3 4B | 43.6 | #1089 |
| 10 | Gemma 3 12B | 39.6 | #671 |
| 11 | Taichu-VLR-3B | 38.9 | #744 |
| 12 | Qwen 2 VL 72B | 37.3 | #460 |
| 13 | URSA-8B | 37.3 | #811 |
| 14 | VLAA-Thinker-Qwen2.5VL-3B | 36.6 | #720 |
| 15 | VLM-R1-3B-Math-0305 | 36.2 | #743 |
Interactive version: theaggregate.ai/benchmark?slug=open-lmm-reasoning-wemath-insufficientknowledge-loose · How It Works · Data refreshed daily, snapshot 2026-10-20.