AMVICC - MMVP Pair Accuracy - Presence of Specific Features: leaderboard

Metric: Pair accuracy (%; the 35 MMVP image pairs in the Presence of Specific Features 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: Estimated already saturated. 11 models tracked.

Top models

#ModelScore
1Llama 3.2 90B Vision Instruct94.29
2Llama 4 Scout91.43
3GPT-4o82.86
4Llama 4 Maverick80
5Gemma 3 27B80
6Qwen 2.5 VL 72B Instruct80
7Gemini 2.5 Pro77.14
8pixtral-large-241177.14
9Claude Sonnet 474.29
10Claude Opus 4.174.29
11Grok 471.43

Interactive version: theaggregate.ai/benchmark?slug=amvicc-mmvp-pair-accuracy-presence-of-specific-features · How It Works · Data refreshed daily, snapshot 2026-09-29.