LLaVA-v1.6-34B: benchmark results
Provider: LLaVA Team. Access: Open.
Aggregate Rating 1412 ± 37, rank #5554 of 10916 rated models, from 43 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FoodSense - Sound | 15.9 | Sound: Lin's concordance correlation coefficient between the | 100 |
| GeoHeight-Bench+ - Elevation Retrieval | 11.32 | Text accuracy (%) on the Elevation Retrieval task (the model | 100 |
| DefectBench (Building Facades) - Crack Counting Error | 0.68 | Mean absolute error of the model's count of crack instances | 94.1 |
| GeoHeight-Bench - Threshold Segmentation | 45.84 | Text accuracy (%) on the Threshold Segmentation task (the mo | 91.7 |
| GeoHeight-Bench - Elevation Retrieval | 51.37 | Text accuracy (%) on the Elevation Retrieval task (the model | 83.3 |
| GeoHeight-Bench+ | 15.42 | Text accuracy (%), the mean of the 8 task accuracies, on Geo | 83.3 |
| GeoHeight-Bench+ - Instance Extraction | 7.27 | Text accuracy (%) on the Instance Extraction task (the model | 83.3 |
| BLINK | 45.05 | Score (%) | 77.8 |
| DefectBench (Building Facades) - External Fixings Counting Error | 0.32 | Mean absolute error of the model's count of external-fixing | 76.5 |
| FoodSense - Texture | 11.3 | Texture: Lin's concordance correlation coefficient between t | 75 |
| GeoHeight-Bench | 21.88 | Text accuracy (%), the mean of the 7 task accuracies, on Geo | 75 |
| GeoHeight-Bench+ - Pixel Information | 11.38 | Text accuracy (%) on the Pixel Information task (the model g | 75 |
Interactive version: theaggregate.ai/model?slug=llava-v1-6-34b · How It Works · Data refreshed daily, snapshot 2026-11-08.