InternVL3-14B: benchmark results
OpenGVLab/Shanghai AI Lab's open 14B vision-language model with native multimodal pretraining (April 2025). Provider: Shanghai AI Lab. Released 2025-04-11. Access: Open.
Aggregate Rating 1582 ± 6, rank #506 of 1635 rated models, from 1290 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MEGA-Bench Task - Next Action Prediction | 85.7 | Task Score (%) | 100 |
| MEGA-Bench Task - OCR Resume Experience Plain | 78.6 | Task Score (%) | 100 |
| MUSE - Relative Position | 10 | Accuracy (%) | 100 |
| OpenVLM ChartQA Test - Augmented | 95.1 | Relaxed Accuracy (%) | 100 |
| OpenVLM MME - Code Reasoning | 200 | Score | 100 |
| RetinaGroundBench - VQA Recall | 76 | Recall (%) | 100 |
| StylisticBias - Age Disparity | 0.04 | Variation strength (0-0.5; between-group dispersion of the f | 100 |
| OpenVLM MMT-Bench - DU | 92.9 | Score (%) | 99.8 |
| OpenVLM MMT-Bench - Image-to-Image Retrieval | 90 | Score (%) | 99.8 |
| OpenVLM MMT-Bench - VPU | 95 | Score (%) | 99.5 |
| OpenVLM MMT-Bench - Vehicle Retrieval | 95 | Score (%) | 99.5 |
| OpenVLM MMT-Bench - VP | 78.3 | Score (%) | 99.3 |
Interactive version: theaggregate.ai/model?slug=internvl3-14b · How It Works · Data refreshed daily, snapshot 2026-11-05.