Open LMM Reasoning - WeMath - RoteMemorization (Loose): leaderboard
Metric: Accuracy (%). Source: huggingface.co. 83 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 Mini | 91.7 |
| 2 | Llama 3.2 11B Instruct | 36 |
| 3 | Qwen 2 VL 2B | 22.3 |
| 4 | InternVL2.5-2B | 19.3 |
| 5 | Aquila-VL-2B | 13.5 |
| 6 | Qwen 2 VL 7B | 12.5 |
| 7 | Gemma 3 4B | 10.3 |
| 8 | GPT-4.1 Nano | 7.8 |
| 9 | InternVL2.5-8B-BoN-8 | 7.8 |
| 10 | URSA-8B | 7.5 |
| 11 | VLM-R1-3B-Math-0305 | 7.5 |
| 12 | VLAA-Thinker-Qwen2.5VL-3B | 6.9 |
| 13 | Grok 2 (1212) | 6.5 |
| 14 | Gemma 3 12B | 6.2 |
| 15 | Taichu-VLR-3B | 5.8 |
Interactive version: theaggregate.ai/benchmark?slug=open-lmm-reasoning-wemath-rotememorization-loose · How It Works · Data refreshed daily, snapshot 2026-09-05.