AMVICC - MMVP Individual Accuracy - Orientation and Direction: leaderboard
AMVICC (Basappa, Goel et al., 2026) asks vision-language models the original MMVP questions about pairs of nearly identical images, grading free-form answers with GPT-4 against the authors' rubric. This board is accuracy in percent on the 30 questions of MMVP's Orientation and Direction category (15 image pairs), each question scored on its own; higher is better.
Metric: Individual accuracy (%; the 30 MMVP questions on 15 image pairs in the Orientation and Direction category, free-form answers graded by GPT-4 against the authors' rubric). Source: arxiv.org. Saturation forecast: Not forecast. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.2 90B Vision Instruct | 90 | #1749 |
| 2 | Gemini 2.5 Pro | 86.67 | #266 |
| 3 | Llama 4 Maverick | 86.67 | #1099 |
| 4 | GPT-4o | 83.33 | #741 |
| 5 | Claude Opus 4.1 | 83.33 | #272 |
| 6 | Qwen 2.5 VL 72B Instruct | 73.33 | #846 |
| 7 | Llama 4 Scout | 70 | #1776 |
| 8 | Gemma 3 27B | 66.67 | #1631 |
| 9 | pixtral-large-2411 | 66.67 | #1309 |
| 10 | Claude Sonnet 4 | 60 | #400 |
| 11 | Grok 4 | 40 | #307 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=amvicc-mmvp-individual-accuracy-orientation-and-direction · How It Works · Data refreshed daily, snapshot 2026-11-13.