InternLM-XComposer: benchmark results
Provider: Shanghai AI Lab. Released 2023-09-26.
Aggregate Rating 1219 ± 46, rank #9488 of 10916 rated models, from 408 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM MMT-Bench - Action Quality Assessment | 50 | Score (%) | 98.8 |
| OpenVLM MME - Scene | 167.5 | Score | 97 |
| OpenVLM MMBench V1.0 CN - RR | 83.4 | Accuracy (%) | 92.6 |
| OpenVLM COCO Captions | 41 | CIDEr | 89 |
| OpenVLM SEEDBench IMG - Instance Interaction | 77.3 | Accuracy (%) | 86.9 |
| OpenVLM MMMU - Architecture and Engineering | 40 | Accuracy (%) | 73.9 |
| OpenVLM CCBench - Cultural Relics | 52.6 | Accuracy (%) | 73.1 |
| OpenVLM MME - Celebrity | 150.6 | Score | 67.5 |
| OpenVLM MMMU - Mechanical Engineering | 40 | Accuracy (%) | 64 |
| OpenVLM CCBench - Scenery and Buildings | 57.9 | Accuracy (%) | 63.9 |
| OpenVLM MMBench V1.0 EN - AR | 82 | Accuracy (%) | 61.3 |
| OpenVLM ScienceQA Test | 89.8 | Accuracy (%) | 60.8 |
Interactive version: theaggregate.ai/model?slug=internlm-xcomposer · How It Works · Data refreshed daily, snapshot 2026-11-08.