AMVICC - MMVP Individual Accuracy - Quantity and Count: leaderboard
Metric: Individual accuracy (%; the 24 MMVP questions on 12 image pairs in the Quantity and Count 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 4 Maverick | 95.83 |
| 2 | Llama 3.2 90B Vision Instruct | 83.33 |
| 3 | Gemini 2.5 Pro | 79.17 |
| 4 | Llama 4 Scout | 79.17 |
| 5 | Qwen 2.5 VL 72B Instruct | 79.17 |
| 6 | pixtral-large-2411 | 79.17 |
| 7 | GPT-4o | 75 |
| 8 | Claude Opus 4.1 | 75 |
| 9 | Claude Sonnet 4 | 70.83 |
| 10 | Gemma 3 27B | 70.83 |
| 11 | Grok 4 | 50 |
Interactive version: theaggregate.ai/benchmark?slug=amvicc-mmvp-individual-accuracy-quantity-and-count · How It Works · Data refreshed daily, snapshot 2026-09-29.