DeepSeek-VL2-Small: benchmark results
Provider: DeepSeek. Released 2024-12-13. Access: Open.
Aggregate Rating 1491 ± 15, rank #2498 of 10916 rated models, from 301 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| PaveVQA - Localization Token-F1 | 40.36 | Token-level F1 (%) of short spatial descriptions on localiza | 100 |
| OpenVLM MMBench V1.0 EN - CP | 89.1 | Accuracy (%) | 99.3 |
| OpenVLM MMBench V1.1 EN - Attribute Recognition | 97.8 | Accuracy (%) | 98.1 |
| OpenVLM MMBench V1.1 EN - Image Style | 87 | Accuracy (%) | 96.5 |
| OpenVLM MME - Celebrity | 177.6 | Score | 96.2 |
| OpenVLM MMBench V1.0 CN - CP | 88.4 | Accuracy (%) | 95.6 |
| OpenVLM MMBench V1.1 CN - Image Style | 83.7 | Accuracy (%) | 95.1 |
| OpenVLM SEEDBench IMG - Instance Attributes | 79.6 | Accuracy (%) | 94 |
| OpenVLM SEEDBench IMG - Instances Counting | 71.9 | Accuracy (%) | 93.1 |
| OpenVLM MMBench V1.1 CN - Image Topic | 98.9 | Accuracy (%) | 90.3 |
| OpenVLM MME - Existence | 200 | Score | 89.7 |
| SEEDBench IMG | 76.9 | Accuracy (%) | 89.3 |
Interactive version: theaggregate.ai/model?slug=deepseek-vl2-small · How It Works · Data refreshed daily, snapshot 2026-11-08.