Chameleon-7B: benchmark results
Provider: Meta. Released 2024-06-16. Access: Open.
Aggregate Rating 1189 ± 48, rank #9458 of 10489 rated models, from 420 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM MMT-Bench - Pixel Localization | 40 | Score (%) | 84.5 |
| OpenVLM MMT-Bench - Raven Progressive Matrices | 20 | Score (%) | 76.2 |
| OpenVLM MMT-Bench - Table Structure Recognition | 75 | Score (%) | 74.5 |
| OpenVLM COCO Captions | 2.1 | CIDEr | 68.3 |
| OpenVLM MMT-Bench - Sketch-to-Code | 30 | Score (%) | 66.5 |
| OpenVLM MMT-Bench - Animal Keypoint Detection | 35 | Score (%) | 65.3 |
| OpenVLM MathVista - LOG | 21.6 | Accuracy (%) | 58.2 |
| OpenVLM MMT-Bench - Rotated Object Detection | 33.3 | Score (%) | 55.1 |
| OpenVLM MMT-Bench - Spot The Diff | 25 | Score (%) | 51.7 |
| Med-RewardBench | 56.15 | Expert-Agreement Accuracy (%; pairwise, six dimensions) | 51.5 |
| OpenVLM MMT-Bench - Image Quality Assessment | 35 | Score (%) | 45.4 |
| OpenVLM MMT-Bench - 3D Indoor Recognition | 25 | Score (%) | 39.3 |
Interactive version: theaggregate.ai/model?slug=chameleon-7b · How It Works · Data refreshed daily, snapshot 2026-11-20.