PaliGemma-3B-mix-448: benchmark results
Provider: Google. Released 2024-05-14. Access: Open.
Aggregate Rating 1376 ± 27, rank #6975 of 10916 rated models, from 582 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM VHELM - Gqa | 74.8 | PEM | 100 |
| OpenVLM MMT-Bench - Pixel Recognition | 80 | Score (%) | 99.3 |
| HELM VHELM - Viz Wiz | 83 | PEM | 97.9 |
| HELM VHELM - Blink Perception - Art Style | 100 | Prefix Quasi-Exact Match (%) | 96.8 |
| HELM VHELM - Blink Reasoning - Jigsaw | 100 | Prefix Quasi-Exact Match (%) | 96.8 |
| OpenVLM COCO Captions | 80.6 | CIDEr | 96.2 |
| OpenVLM MMT-Bench - Human Interaction Understanding | 90 | Score (%) | 95.9 |
| HELM VHELM - Blink Perception - Relative Reflectance | 68.66 | Prefix Quasi-Exact Match (%) | 95.7 |
| OpenVLM MMT-Bench - Gaze Estimation | 35 | Score (%) | 93 |
| OpenVLM MMT-Bench - Food Recognition | 100 | Score (%) | 92.5 |
| OpenVLM MMT-Bench - Geometrical Perspective | 83.3 | Score (%) | 90.5 |
| OpenVLM MMT-Bench - Image Matting | 30 | Score (%) | 90.3 |
Interactive version: theaggregate.ai/model?slug=paligemma-3b-mix-448 · How It Works · Data refreshed daily, snapshot 2026-11-08.