InternVL2.5-2B-MPO: benchmark results
Provider: Shanghai AI Lab. Released 2024-12-20. Access: Open.
Aggregate Rating 1395 ± 20, rank #6325 of 10916 rated models, from 518 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM CCBench - Food and Clothes | 90.4 | Accuracy (%) | 100 |
| OpenVLM SEEDBench IMG - Text Understanding | 85.7 | Accuracy (%) | 100 |
| CCBench | 86.1 | Accuracy (%) | 99.6 |
| OpenVLM CCBench - Traditional Shows | 89.4 | Accuracy (%) | 99.6 |
| OpenVLM CCBench - Calligraphy and Painting | 91.2 | Accuracy (%) | 99.3 |
| OpenVLM CCBench - Cultural Relics | 92.8 | Accuracy (%) | 99.1 |
| OpenVLM CCBench - Sketch Reasoning | 95.6 | Accuracy (%) | 98.7 |
| OpenVLM MMT-Bench - Multiple View Image Understanding | 50 | Score (%) | 98.1 |
| OpenVLM MMT-Bench - Image Matting | 40 | Score (%) | 97.8 |
| OpenVLM POPE | 90 | Overall (%) | 97.7 |
| Open LMM Reasoning - WeMath - RoteMemorization (Strict) | 84.6 | Accuracy (%) | 97.5 |
| OpenVLM CCBench - Scenery and Buildings | 92.6 | Accuracy (%) | 97.4 |
Interactive version: theaggregate.ai/model?slug=internvl2-5-2b-mpo · How It Works · Data refreshed daily, snapshot 2026-11-08.