Nemotron 3 Nano Omni 30B (Reasoning) vs Claude Opus 5 (high)
Results updated 15 November 2026 · In Terminus-2 and Claude Code, so part of any gap can come from the harness · How we benchmark
Verdict
Claude Opus 5 (high) clearly beats Nemotron 3 Nano Omni 30B (Reasoning) by 59 points, and is faster.
- Cost per task
- No data vs $0.32
- Time per task
- 4.6 min vs 1.1 min
- Tokens per attempt
- 564k vs 173k
- Image input
- Yes vs Yes
Claude Opus 5 (high), Claude Code scores higher than Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2 on all 3 evaluations scored for both.
Where they differ most
By kind of work, share of tasks passed, 0 to 100 · Higher is better
| Kind of work | Nemotron 3 Nano Omni 30B (Reasoning) | Claude Opus 5 (high) |
|---|---|---|
| GIS analysis | 7 | 88 |
| Remote sensing and raster analysis | 17 | 85 |
| Making maps | 16 | 82 |
| Climate and time series | 28 | 80 |
Charts for each: Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2 · Claude Opus 5 (high), Claude Code.
Side by side
| Metric | Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2 | Claude Opus 5 (high), Claude Code |
|---|---|---|
| Geospatial Agent Index: Text only | 21 | 80 |
| Geospatial Agent Index: Text + vision | No data | 81 |
| Earth-Bench V2 score | 31 | 74 |
| GeoBenchX score | 23 | 75 |
| GS-QA2 score | 9 | 91 |
| Vision agents score | No data | 84 |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 28 | 100 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 16 | 89 |
| OGD4All score (standalone; in no index) | 47 | 95 |
| GeoNatureAgent score (standalone; in no index) | 17 | 100 |
| MultiGlobeQA score (standalone; in no index) | 48 | 79 |
| GeoAgentEval score (standalone; in no index) | 48 | 96 |
| ImpactMesh score (standalone; in no index) | No data | No data |
| OpenEarthAgent Maps score (standalone; in no index) | 33 | 99 |
| OpenEarthAgent Sentinel-2 score (standalone; in no index) | 52 | 100 |
| GS-QA2 Full score (standalone; in no index) | No data | 86 |
| Place and coordinate tasks score (standalone; in no index) | 60 | No data |
| InSAR Regional Snow Mapping score (standalone; in no index) | 33 | 100 |
| DLPU phase unwrapping score (standalone; in no index) | No data | No data |
| Remote sensing and raster analysis score | 17 | 85 |
| Change detection score | 28 | 68 |
| GIS analysis score | 7 | 88 |
| Climate and time series score | 28 | 80 |
| Making maps score | 16 | 82 |
| Cost per task | No data | $0.32 |
| Time per task (at 100 output tokens/s) | 4.6 min | 1.1 min |
| Tokens per attempt | 564k | 173k |
| Turns per attempt | 30.2 | 7.2 |
| Input price per 1M tokens | No data | $5.00 |
| Output price per 1M tokens | No data | $25.00 |
| Context window | 256,000 | 1,000,000 |
| Image input | Yes | Yes |
| Developer | NVIDIA | Anthropic |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2 or Claude Opus 5 (high), Claude Code?
Claude Opus 5 (high), Claude Code, 80 to 21.
Which is cheaper per task, Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2 or Claude Opus 5 (high), Claude Code?
Not known yet: Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2 has no cost data.
Which writes fewer output tokens per task, Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2 or Claude Opus 5 (high), Claude Code?
Claude Opus 5 (high), Claude Code, 4k to 21k.
Which has the larger context window, Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2 or Claude Opus 5 (high), Claude Code?
Claude Opus 5 (high), Claude Code, 1,000,000 tokens to 256,000 tokens.
