Ovis2.5-2B: benchmark results
Provider: Alibaba. Released 2025-08-15. Access: Open.
Aggregate Rating 1544 ± 26, rank #1775 of 10916 rated models, from 51 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MathVista | 79.8 | Accuracy (%) | 96.2 |
| HallusionBench | 58.5 | Overall Score | 88.5 |
| MMStar | 67.7 | Accuracy (%) | 82.2 |
| OpenVLM OCRBench | 86.8 | Score (normalized) | 82.1 |
| MM-Vet | 68.3 | Score | 78.7 |
| AI2D | 85 | Accuracy (%) | 77.9 |
| MCSBench - L3 | 25.19 | Accuracy (%) | 65.1 |
| OpenCompass Multimodal - Average | 73 | Average Score | 63.5 |
| VISTA-Bench - Multimodal Perception | 66.3 | Accuracy (%) on the 300 multimodal perception questions (glo | 62.1 |
| OpenCompass Multimodal - MMMU | 58.7 | Score (%) | 59.1 |
| VISTA-Bench - Multimodal Reasoning | 51.7 | Accuracy (%) on the 300 multimodal reasoning questions (logi | 58.6 |
| TSHA - Open-Ended Hazard QA | 71.7 | Open-ended QA overall (%): GPT-4o rubric score, 0.7 hazard a | 57.1 |
Interactive version: theaggregate.ai/model?slug=ovis2-5-2b · How It Works · Data refreshed daily, snapshot 2026-11-08.