A.X-4.0-VL-Light: benchmark results
Provider: SK Telecom. Access: Open.
Aggregate Rating 1470 ± 27, rank #2841 of 10916 rated models, from 19 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| K-MetBench | 52.5 | Accuracy (imported) | 41.4 |
| KorMedMCQA-V - Three or More Images | 41.9 | Accuracy (%) on the 39 questions with three or more images ( | 38 |
| KorMedMCQA-V - Ultrasound | 43.4 | Accuracy (%) on the questions carrying the 138 ultrasound im | 36 |
| KorMedMCQA-V - Non-Stress Test | 39.9 | Accuracy (%) on the questions carrying the 54 fetal non-stre | 28 |
| KorMedMCQA-V - Two Images | 38.7 | Accuracy (%) on the 426 questions with two images (KorMedMCQ | 26 |
| KorMedMCQA-V - MRI | 44.4 | Accuracy (%) on the questions carrying the 40 MRI images (Ko | 25 |
| PetQA-Bench - Multimodal Factuality | 46.1 | Factuality (%) | 25 |
| KorMedMCQA-V - CT | 37 | Accuracy (%) on the questions carrying the 336 CT images (Ko | 24 |
| KorMedMCQA-V | 41.8 | Accuracy (%) on all 1,534 questions (KorMedMCQA-V: image-bas | 23 |
| KorMedMCQA-V - ECG | 43.5 | Accuracy (%) on the questions carrying the 164 electrocardio | 22 |
| KorMedMCQA-V - Single Image | 43.1 | Accuracy (%) on the 1,069 questions with one image (KorMedMC | 22 |
| KorMedMCQA-V - Endoscopy | 35.5 | Accuracy (%) on the questions carrying the 122 endoscopy ima | 20 |
Interactive version: theaggregate.ai/model?slug=a-x-4-0-vl-light · How It Works · Data refreshed daily, snapshot 2026-11-08.