AMVICC - MMVP Individual Accuracy - Presence of Specific Features: leaderboard
Metric: Individual accuracy (%; the 70 MMVP questions on 35 image pairs in the Presence of Specific Features category, free-form answers graded by GPT-4 against the authors' rubric). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.2 90B Vision Instruct | 97.14 |
| 2 | Llama 4 Scout | 95.71 |
| 3 | GPT-4o | 91.43 |
| 4 | Llama 4 Maverick | 90 |
| 5 | Gemma 3 27B | 90 |
| 6 | Qwen 2.5 VL 72B Instruct | 90 |
| 7 | Gemini 2.5 Pro | 88.57 |
| 8 | pixtral-large-2411 | 88.57 |
| 9 | Claude Sonnet 4 | 87.14 |
| 10 | Claude Opus 4.1 | 87.14 |
| 11 | Grok 4 | 82.86 |
Interactive version: theaggregate.ai/benchmark?slug=amvicc-mmvp-individual-accuracy-presence-of-specific-features · How It Works · Data refreshed daily, snapshot 2026-09-29.