Mistral Large 4 (high) vs Step 5 Preview (Reasoning)
Results updated 3 November 2026 · Both in Terminus-2 · How we benchmark
Verdict
Mistral Large 4 (high) beats Step 5 Preview (Reasoning) by 8 points, and is cheaper and faster.
- Cost per task
- $0.10 vs $0.13
- Time per task
- 12.0 min vs 13.1 min (estimated)
- Tokens per attempt
- 226k vs 311k
- Image input
- Yes vs Yes
Mistral Large 4 (high), Terminus-2 scores higher on GeoBenchX; Step 5 Preview (Reasoning), Terminus-2 scores higher on Earth-Bench V2 and GS-QA2. Mistral Large 4 (high), Terminus-2 is cheaper and faster: $0.10 against $0.13 per task, and 12.0 min against 13.1 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) | Step 5 Preview (Reasoning) |
|---|---|---|
| Making maps | 82 | 21 |
| Spatial statistics | 46 | 0 |
| Climate and time series | 40 | 54 |
| Remote sensing and raster analysis | 67 | 62 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: Mistral Large 4 (high), Terminus-2 · Step 5 Preview (Reasoning), Terminus-2.
Side by side
| Metric | Mistral Large 4 (high), Terminus-2 | Step 5 Preview (Reasoning), Terminus-2 |
|---|---|---|
| Geospatial Agent Index: Text only | 59 | 51 |
| Geospatial Agent Index: Text + vision | No data | No data |
| Earth-Bench V2 score | 46 | 56 |
| GeoBenchX score | 55 | 18 |
| GS-QA2 score | 77 | 79 |
| Vision agents score | No data | No data |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 80 | 84 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 94 | 89 |
| OGD4All score (standalone; in no index) | 96 | 93 |
| GeoNatureAgent score (standalone; in no index) | 100 | 88 |
| MultiGlobeQA score (standalone; in no index) | 62 | 70 |
| GeoAgentEval score (standalone; in no index) | 60 | 91 |
| Remote sensing and raster analysis score | 67 | 62 |
| GIS analysis score | 82 | 82 |
| Climate and time series score | 40 | 54 |
| Spatial statistics score | 46 | 0 |
| Making maps score | 82 | 21 |
| Recognising infeasible requests score | 23 | 23 |
| Cost per task | $0.10 | $0.13 |
| Time per task (at 100 output tokens/s) | 12.0 min | 13.1 min (estimated) |
| Tokens per attempt | 226k | 311k |
| Turns per attempt | 13.8 | 15.1 |
| Input price per 1M tokens | $0.70 | $1.00 |
| Output price per 1M tokens | $2.10 | $2.70 |
| Context window | 262,144 | 1,024,000 |
| Image input | Yes | Yes |
| Developer | Mistral AI | StepFun |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, Mistral Large 4 (high), Terminus-2 or Step 5 Preview (Reasoning), Terminus-2?
Mistral Large 4 (high), Terminus-2, 59 to 51.
Which is cheaper per task, Mistral Large 4 (high), Terminus-2 or Step 5 Preview (Reasoning), Terminus-2?
Mistral Large 4 (high), Terminus-2, $0.10 to $0.13.
Which writes fewer output tokens per task, Mistral Large 4 (high), Terminus-2 or Step 5 Preview (Reasoning), Terminus-2?
Mistral Large 4 (high), Terminus-2, 14k to 24k.
Which has the larger context window, Mistral Large 4 (high), Terminus-2 or Step 5 Preview (Reasoning), Terminus-2?
Step 5 Preview (Reasoning), Terminus-2, 1,024,000 tokens to 262,144 tokens.