Mistral Large 4 (high) vs Nemotron 3 Ultra (high)
Results updated 3 November 2026 · Both in Terminus-2 · How we benchmark
Verdict
Nemotron 3 Ultra (high) beats Mistral Large 4 (high) by 9 points, and is cheaper and faster.
- Cost per task
- $0.10 vs $0.067
- Time per task
- 12.0 min vs 2.4 min
- Tokens per attempt
- 226k vs 144k
- Image input
- Yes vs No
Mistral Large 4 (high), Terminus-2 scores higher on GS-QA2; Nemotron 3 Ultra (high), Terminus-2 scores higher on Earth-Bench V2 and GeoBenchX. Nemotron 3 Ultra (high), Terminus-2 is cheaper and faster: $0.067 against $0.10 per task, and 2.4 min against 12.0 min.
Where they differ most
By kind of work, share of tasks passed, 0 to 100 · Higher is better
| Kind of work | Mistral Large 4 (high) | Nemotron 3 Ultra (high) |
|---|---|---|
| Recognising infeasible requests | 23 | 68 |
| Climate and time series | 40 | 68 |
| Spatial statistics | 46 | 31 |
| Making maps | 82 | 70 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: Mistral Large 4 (high), Terminus-2 · Nemotron 3 Ultra (high), Terminus-2.
Side by side
| Metric | Mistral Large 4 (high), Terminus-2 | Nemotron 3 Ultra (high), Terminus-2 |
|---|---|---|
| Geospatial Agent Index: Text only | 59 | 68 |
| Geospatial Agent Index: Text + vision | No data | No data |
| Earth-Bench V2 score | 46 | 69 |
| GeoBenchX score | 55 | 65 |
| GS-QA2 score | 77 | 70 |
| Vision agents score | No data | No data |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 80 | 96 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 94 | 63 |
| OGD4All score (standalone; in no index) | 96 | 92 |
| GeoNatureAgent score (standalone; in no index) | 100 | 92 |
| MultiGlobeQA score (standalone; in no index) | 62 | 72 |
| GeoAgentEval score (standalone; in no index) | 60 | 92 |
| Remote sensing and raster analysis score | 67 | 70 |
| GIS analysis score | 82 | 70 |
| Climate and time series score | 40 | 68 |
| Spatial statistics score | 46 | 31 |
| Making maps score | 82 | 70 |
| Recognising infeasible requests score | 23 | 68 |
| Cost per task | $0.10 | $0.067 |
| Time per task (at 100 output tokens/s) | 12.0 min | 2.4 min |
| Tokens per attempt | 226k | 144k |
| Turns per attempt | 13.8 | 12.0 |
| Input price per 1M tokens | $0.70 | $0.50 |
| Output price per 1M tokens | $2.10 | $2.20 |
| Context window | 262,144 | 262,144 |
| Image input | Yes | No |
| Developer | Mistral AI | NVIDIA |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, Mistral Large 4 (high), Terminus-2 or Nemotron 3 Ultra (high), Terminus-2?
Nemotron 3 Ultra (high), Terminus-2, 68 to 59.
Which is cheaper per task, Mistral Large 4 (high), Terminus-2 or Nemotron 3 Ultra (high), Terminus-2?
Nemotron 3 Ultra (high), Terminus-2, $0.067 to $0.10.
Which writes fewer output tokens per task, Mistral Large 4 (high), Terminus-2 or Nemotron 3 Ultra (high), Terminus-2?
Nemotron 3 Ultra (high), Terminus-2, 9k to 14k.
Which has the larger context window, Mistral Large 4 (high), Terminus-2 or Nemotron 3 Ultra (high), Terminus-2?
They are level, at 262,144 tokens.