Ovis2-16B: benchmark results
Provider: Alibaba. Released 2025-02-21. Access: Open.
Aggregate Rating 1557 ± 13, rank #1594 of 10916 rated models, from 597 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM MMBench V1.0 CN - FP-C | 89.9 | Accuracy (%) | 100 |
| OpenVLM MMBench V1.1 CN - Attribute Comparison | 96 | Accuracy (%) | 100 |
| OpenVLM MMBench V1.1 EN - Attribute Comparison | 97.3 | Accuracy (%) | 100 |
| OpenVLM MMT-Bench - DocVQA | 95 | Score (%) | 100 |
| OpenVLM Q-Bench - What (Others) | 89.3 | Accuracy (%) | 100 |
| VLM-DeflectionBench - Realistic Accuracy | 61.1 | Accuracy (%) of knowledge-based visual question answering in | 100 |
| OpenVLM MMBench V1.0 EN - AR | 89.2 | Accuracy (%) | 99.6 |
| OpenVLM SEED-Bench 2 Plus - Maps | 64.3 | Accuracy (%) | 99.5 |
| OpenVLM A-Bench - Contradiction Overcome | 95.8 | Accuracy (%) | 99.4 |
| OpenVLM MMT-Bench - Multiple Instance Captioning | 100 | Score (%) | 99.3 |
| OpenVLM MMT-Bench - Polygon Localization | 75 | Score (%) | 99.3 |
| OpenVLM MMBench V1.0 CN - AR | 88.9 | Accuracy (%) | 99.1 |
Interactive version: theaggregate.ai/model?slug=ovis2-16b · How It Works · Data refreshed daily, snapshot 2026-11-08.