AMVICC - MMVP Individual Accuracy - Text: leaderboard
Metric: Individual accuracy (%; the 14 MMVP questions on 7 image pairs in the Text category, free-form answers graded by GPT-4 against the authors' rubric). Source: arxiv.org. Saturation forecast: Around December 2026. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 92.86 |
| 2 | Llama 4 Maverick | 92.86 |
| 3 | GPT-4o | 85.71 |
| 4 | Llama 4 Scout | 85.71 |
| 5 | Llama 3.2 90B Vision Instruct | 85.71 |
| 6 | Qwen 2.5 VL 72B Instruct | 85.71 |
| 7 | Claude Opus 4.1 | 85.71 |
| 8 | Claude Sonnet 4 | 78.57 |
| 9 | Gemma 3 27B | 78.57 |
| 10 | pixtral-large-2411 | 71.43 |
| 11 | Grok 4 | 64.29 |
Interactive version: theaggregate.ai/benchmark?slug=amvicc-mmvp-individual-accuracy-text · How It Works · Data refreshed daily, snapshot 2026-09-29.