NVILA-15B: benchmark results
Provider: Other. Access: Open.
Aggregate Rating 1523 ± 40, rank #2035 of 10916 rated models, from 35 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| PhysBench | 46.9 | Overall Accuracy (%) | 85 |
| MCSBench - L1 | 67.82 | Accuracy (%) | 67.8 |
| GeoMMBench - SAR | 49.7 | Micro-averaged accuracy (%) on the synthetic aperture radar | 64.7 |
| GST-Bench - Top-Down Selection (Hard) | 32.02 | Accuracy (%; multiple choice of the top-down image of the ex | 61.9 |
| GeoMMBench - Photogrammetry | 62.2 | Micro-averaged accuracy (%) on the photogrammetry discipline | 61.8 |
| MMSI-Bench | 30.5 | Accuracy (%) | 60.5 |
| GeoMMBench - Applications | 57.6 | Micro-averaged accuracy (%) on the remote sensing applicatio | 60.3 |
| GeoMMBench - Thermal Imagery | 50 | Micro-averaged accuracy (%) on the thermal imagery questions | 60.3 |
| GeoMMBench - DEM | 47.3 | Micro-averaged accuracy (%) on the digital elevation model q | 58.8 |
| MCSBench - L3 | 24.43 | Accuracy (%) | 58.6 |
| GeoMMBench - Principles | 48.3 | Micro-averaged accuracy (%) on the principles and theoretica | 55.9 |
| GeoMMBench - Multispectral and Hyperspectral | 56.8 | Micro-averaged accuracy (%) on the multispectral and hypersp | 52.9 |
Interactive version: theaggregate.ai/model?slug=nvila-15b · How It Works · Data refreshed daily, snapshot 2026-11-08.