LLaVA-NeXT-34B: benchmark results
Provider: LLaVA Team. Released 2024-01-30. Access: Open.
Aggregate Rating 1459 ± 24, rank #3371 of 10916 rated models, from 56 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| GeoMMBench - Spatial | 59.5 | Micro-averaged accuracy (%) on the spatial relation question | 92.6 |
| GeoMMBench - LiDAR | 74.2 | Micro-averaged accuracy (%) on the LiDAR questions of the Ge | 80.9 |
| MRR-Benchmark | 412 | Total Column Score | 76.9 |
| GeoMMBench - SAR | 52.2 | Micro-averaged accuracy (%) on the synthetic aperture radar | 69.1 |
| GeoMMBench - Time Series | 54.3 | Micro-averaged accuracy (%) on the time-series questions of | 67.6 |
| EIBench (complex) | 35.37 | Recall | 66.7 |
| MUSE - Activity Description | 36.5 | Accuracy (%) | 65.5 |
| VG-LLM Release v3 - CVBench2D | 73 | Accuracy (%) | 65 |
| GeoMMBench - Photogrammetry | 63 | Micro-averaged accuracy (%) on the photogrammetry discipline | 64.7 |
| GeoMMBench - Principles | 50.7 | Micro-averaged accuracy (%) on the principles and theoretica | 61.8 |
| GeoMMBench | 54.2 | Micro-averaged accuracy (%) on all questions of the GeoMMBen | 58.8 |
| GeoMMBench - Remote Sensing | 52.8 | Micro-averaged accuracy (%) on the remote sensing discipline | 57.4 |
Interactive version: theaggregate.ai/model?slug=llava-next-34b · How It Works · Data refreshed daily, snapshot 2026-11-08.