OpenVLM Subjective — leaderboard
Metric: Average Normalized Score (%). Source: huggingface.co. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 (2025-04-14) | 76.6 |
| 2 | Gemini 2.5 Flash | 71.73 |
| 3 | Claude 3.7 Sonnet | 71.47 |
| 4 | Gemini 2.0 Flash | 70.11 |
| 5 | Grok 2 (1212) | 68.91 |
| 6 | Claude 3.5 Sonnet (20241022) | 68.9 |
| 7 | InternVL3-78B | 67.61 |
| 8 | Llama 4 Scout Instruct | 65.54 |
| 9 | InternVL3-38B | 64.31 |
| 10 | InternVL3-8B | 63.66 |
| 11 | InternVL2.5-78B | 62.7 |
| 12 | Pixtral-12B | 60.42 |
| 13 | Llama 3.2 11B Instruct | 48.1 |
Interactive version: aibenchmarks.dev/benchmark?slug=openvlm-subjective · How the rankings work · Data refreshed daily, snapshot 2026-07-13.