Open LMM Reasoning - WeMath - InadequateGeneralization (Loose): leaderboard
Metric: Accuracy (%). Source: huggingface.co. 83 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL3-8B | 16.2 |
| 2 | Gemma 3 4B | 15.6 |
| 3 | Taichu-VLR-3B | 15.4 |
| 4 | Claude 3.5 Sonnet (20241022) | 15.2 |
| 5 | Gemma 3 27B | 15.1 |
| 6 | Gemma 3 12B | 14.7 |
| 7 | VLAA-Thinker-Qwen2.5VL-7B | 14.5 |
| 8 | Qwen 2 VL 72B | 13.7 |
| 9 | VLAA-Thinker-Qwen2.5VL-3B | 13.5 |
| 10 | Qwen 2 VL 7B | 13.3 |
| 11 | VLM-R1-3B-Math-0305 | 13.1 |
| 12 | Gemini 1.5 Pro (002) | 12.9 |
| 13 | InternVL2.5-78B | 12.9 |
| 14 | Gemini 2.0 Flash | 12.9 |
| 15 | GPT-4.1 Nano | 12.8 |
Interactive version: theaggregate.ai/benchmark?slug=open-lmm-reasoning-wemath-inadequategeneralization-loose · How It Works · Data refreshed daily, snapshot 2026-09-05.