InternVL2.5-4B-MPO: benchmark results
Provider: Shanghai AI Lab. Released 2024-12-20. Access: Open.
Aggregate Rating 1467 ± 19, rank #2954 of 10916 rated models, from 496 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM CCBench - Food and Clothes | 87 | Accuracy (%) | 98.7 |
| CCBench | 82.2 | Accuracy (%) | 98.2 |
| OpenVLM MMT-Bench - Handwritten Mathematical Expression Recognition | 100 | Score (%) | 97.8 |
| OpenVLM MMT-Bench - Set-of-Mark Recognition | 100 | Score (%) | 97.8 |
| OpenVLM MMT-Bench - Anatomy Identification | 85 | Score (%) | 97.3 |
| OpenVLM MMT-Bench - Scene Recognition | 85 | Score (%) | 97.3 |
| OpenVLM MMT-Bench - MedU | 83 | Score (%) | 97.1 |
| OpenVLM MME - Text Translation | 200 | Score | 97 |
| OpenVLM CCBench - Cultural Relics | 87.6 | Accuracy (%) | 96.5 |
| OpenVLM CCBench - Traditional Shows | 83.3 | Accuracy (%) | 96.5 |
| OpenVLM MME - Artwork | 164.5 | Score | 95.3 |
| OpenVLM POPE | 89.6 | Overall (%) | 95.3 |
Interactive version: theaggregate.ai/model?slug=internvl2-5-4b-mpo · How It Works · Data refreshed daily, snapshot 2026-11-08.