AMVICC - MMVP Pair Accuracy - Text: leaderboard
Metric: Pair accuracy (%; the 7 MMVP image pairs in the Text category, a pair credited only when both of its questions are answered correctly; free-form answers graded by GPT-4 against the authors' rubric). Source: arxiv.org. Saturation forecast: Around September 2027. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 85.71 |
| 2 | Llama 4 Maverick | 85.71 |
| 3 | GPT-4o | 71.43 |
| 4 | Gemma 3 27B | 71.43 |
| 5 | Llama 4 Scout | 71.43 |
| 6 | Llama 3.2 90B Vision Instruct | 71.43 |
| 7 | Qwen 2.5 VL 72B Instruct | 71.43 |
| 8 | Claude Opus 4.1 | 71.43 |
| 9 | Claude Sonnet 4 | 57.14 |
| 10 | Grok 4 | 42.86 |
| 11 | pixtral-large-2411 | 42.86 |
Interactive version: theaggregate.ai/benchmark?slug=amvicc-mmvp-pair-accuracy-text · How It Works · Data refreshed daily, snapshot 2026-09-29.