GeoChat: benchmark results
Provider: Other. Access: Open.
Aggregate Rating 1490 ± 34, rank #2503 of 10916 rated models, from 105 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| UAVBench - Region Classification (RefDrone) | 53.66 | Region-level classification accuracy (%) of a boxed object, | 100 |
| CVSBench - FOV Ground-to-Satellite | 56.19 | Accuracy (%) on FOV-subset ground-to-satellite questions (li | 92.3 |
| OmniEarth - Change Description | 102.4 | CIDEr against reference descriptions, as the paper prints it | 86.7 |
| OmniEarth - Image Captioning | 138.1 | CIDEr against reference descriptions, as the paper prints it | 86.7 |
| UAVBench - VQA Flying Altitude | 36.51 | Image VQA accuracy (%) on flying-altitude recognition, zero- | 80 |
| UAVBench - Region Classification (WebUAV-3M) | 14.74 | Region-level classification accuracy (%) of a boxed object, | 70 |
| LandmarkBias-3K - Region (200 km) | 38.77 | Accuracy (%) | 69.2 |
| OmniEarth - Generalized Referring Expression Segmentation | 9.5 | Mean intersection over union (%) of predicted masks, General | 66.7 |
| CVSBench - FOV Satellite-to-Ground | 44.44 | Accuracy (%) on FOV-subset satellite-to-ground questions (li | 64.3 |
| OmniEarth - Long-Term Trend Reasoning | 99 | CIDEr against reference descriptions, as the paper prints it | 63.3 |
| RSHR-Bench - Multiple Choice - MTEM@1 | 19.3 | Accuracy (%) | 62.5 |
| UAVBench - Region Classification (DroneVehicle) | 70.51 | Region-level classification accuracy (%) of a boxed object, | 62.5 |
Interactive version: theaggregate.ai/model?slug=geochat · How It Works · Data refreshed daily, snapshot 2026-11-08.