Ovis-U1-3B: benchmark results
Provider: Alibaba. Released 2025-06-28. Access: Open.
Aggregate Rating 1496 ± 26, rank #2417 of 10916 rated models, from 176 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| UniG2U-Bench (Direct) - Real-world Applications | 45.5 | Accuracy (%) on the 200 real-world application items (attent | 100 |
| UniG2U-Bench (Direct) - Puzzles and Games | 27.28 | Accuracy (%) on the 537 puzzle and game items (mental recons | 96.2 |
| OpenVLM AI2D - Food Chains and Webs | 95.8 | Accuracy (%) | 95.8 |
| OpenVLM MMBench V1.1 CN - Image Quality | 66.1 | Accuracy (%) | 95.6 |
| OpenVLM OCRBench - Handwritten Mathematical Expression Recognition | 81 | Correct Answers (%) | 94.5 |
| OpenVLM MMStar - Coarse Perception | 73.2 | Accuracy (%) | 93.3 |
| UniG2U-Bench (Direct) - Chart Reasoning | 90 | Accuracy (%) on the 100 ChartQA items, from UniG2U-Bench, an | 92.3 |
| OpenVLM MMBench V1.1 EN - Image Style | 84.8 | Accuracy (%) | 91.9 |
| OpenVLM OCRBench - Key Information Extraction | 91 | Correct Answers (%) | 90.1 |
| OpenVLM AI2D - Water, Carbon, and Nitrogen Cycle | 75 | Accuracy (%) | 88.7 |
| OpenVLM MMMU - Physics | 66.7 | Accuracy (%) | 88.2 |
| OpenVLM HallusionBench - aAcc | 71.7 | Accuracy (%) | 87.3 |
Interactive version: theaggregate.ai/model?slug=ovis-u1-3b · How It Works · Data refreshed daily, snapshot 2026-11-08.