MMAlignBench — leaderboard
Metric: Overall Reward. Source: huggingface.co. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3.7 Sonnet | 60.32 |
| 2 | GPT-4.1 (2025-04-14) | 59.92 |
| 3 | Grok 2 (1212) | 53.17 |
| 4 | Gemini 2.0 Flash | 40.28 |
| 5 | Claude 3.5 Sonnet (20241022) | 34.92 |
| 6 | Gemini 2.5 Flash | 32.14 |
| 7 | Pixtral-12B | 21.23 |
| 8 | Llama 4 Scout Instruct | 14.29 |
| 9 | InternVL3-78B | -0.79 |
| 10 | InternVL3-8B | -7.94 |
| 11 | InternVL3-38B | -14.68 |
| 12 | InternVL2.5-78B | -19.64 |
| 13 | Llama 3.2 11B Instruct | -37.7 |
Interactive version: aibenchmarks.dev/benchmark?slug=mmalignbench · How the rankings work · Data refreshed daily, snapshot 2026-07-13.