Granite-Vision-3.3-2B: benchmark results
Provider: IBM. Released 2025-06-11. Access: Open.
Aggregate Rating 1372 ± 29, rank #7128 of 10916 rated models, from 478 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM MMBench V1.1 EN - Image Topic | 100 | Accuracy (%) | 100 |
| OpenVLM MMT-Bench - Abstract Visual Recognition | 95 | Score (%) | 100 |
| OpenVLM MMT-Bench - Pixel Localization | 65 | Score (%) | 100 |
| OpenVLM MMMU - Music | 50 | Accuracy (%) | 99.5 |
| OpenVLM MMT-Bench - Existence Hallucination | 100 | Score (%) | 99 |
| OpenVLM COCO Captions | 78.5 | CIDEr | 95.7 |
| OpenVLM MMT-Bench - Logo and Brand Recognition | 100 | Score (%) | 94.4 |
| OpenVLM MMT-Bench - Whoops | 95 | Score (%) | 94.2 |
| OpenVLM MMT-Bench - Deepfake Detection | 65 | Score (%) | 93.7 |
| OpenVLM MMT-Bench - Industrial Produce Anomaly Detection | 85 | Score (%) | 93.7 |
| OpenVLM MMT-Bench - MIA | 75 | Score (%) | 93.2 |
| OpenVLM MMBench V1.1 EN - Identity Reasoning | 98.7 | Accuracy (%) | 91.5 |
Interactive version: theaggregate.ai/model?slug=granite-vision-3-3-2b · How It Works · Data refreshed daily, snapshot 2026-11-08.