NVIDIA models: geospatial agent results
Geospatial Agent Benchmarks tests 3 entries from NVIDIA.
Highlights
- Highest Geospatial Agent Index (ranked entries): Nemotron 3 Ultra (high), Terminus-2 (68).
- Lowest cost per task (ranked entries): Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2 ($0.0096).
- Fewest output tokens per task (ranked entries): Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2 (7k).
Results
Geospatial Agent Index
Equal average of Earth-Bench V2, GeoBenchX and GS-QA2, 0 to 100 · Higher is better
Data table: Geospatial Agent Index
| Model | Geospatial Agent Index: Text only | Earth-Bench V2 | GeoBenchX | GS-QA2 | Vision agents | Cost per task | Time per task (at 100 output tokens/s) | Context window |
|---|---|---|---|---|---|---|---|---|
| Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2 | 20 | 29 | 21 | 9 | No data | $0.13 | 4.9 min | 256,000 |
| Nemotron 3 Ultra (high), Terminus-2 | 68 | 68 | 65 | 70 | No data | $0.069 | 2.4 min | 262,144 |
| Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2 | 52 | 46 | 53 | 59 | No data | $0.0096 | 3.8 min | 262,144 |