Gemma 4 26B A4B (Reasoning) vs Mistral Large 4 (high)
Results updated 3 November 2026 · Both in Terminus-2 · How we benchmark
Verdict
Mistral Large 4 (high) beats Gemma 4 26B A4B (Reasoning) by 3 points; Gemma 4 26B A4B (Reasoning) is cheaper and faster.
- Cost per task
- $0.021 vs $0.10
- Time per task
- 5.6 min vs 12.0 min
- Tokens per attempt
- 168k vs 226k
- Image input
- Yes vs Yes
Gemma 4 26B A4B (Reasoning), Terminus-2 scores higher on GeoBenchX; Mistral Large 4 (high), Terminus-2 scores higher on GS-QA2; they are level on Earth-Bench V2. Gemma 4 26B A4B (Reasoning), Terminus-2 is cheaper and faster: $0.021 against $0.10 per task, and 5.6 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 | Gemma 4 26B A4B (Reasoning) | Mistral Large 4 (high) |
|---|---|---|
| Recognising infeasible requests | 63 | 23 |
| GIS analysis | 57 | 82 |
| Spatial statistics | 31 | 46 |
| Making maps | 70 | 82 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: Gemma 4 26B A4B (Reasoning), Terminus-2 · Mistral Large 4 (high), Terminus-2.
Side by side
| Metric | Gemma 4 26B A4B (Reasoning), Terminus-2 | Mistral Large 4 (high), Terminus-2 |
|---|---|---|
| Geospatial Agent Index: Text only | 56 | 59 |
| Geospatial Agent Index: Text + vision | No data | No data |
| Earth-Bench V2 score | 46 | 46 |
| GeoBenchX score | 63 | 55 |
| GS-QA2 score | 58 | 77 |
| Vision agents score | No data | No data |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 86 | 80 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 63 | 94 |
| OGD4All score (standalone; in no index) | No data | 96 |
| GeoNatureAgent score (standalone; in no index) | No data | 100 |
| MultiGlobeQA score (standalone; in no index) | No data | 62 |
| GeoAgentEval score (standalone; in no index) | No data | 60 |
| Remote sensing and raster analysis score | 59 | 67 |
| GIS analysis score | 57 | 82 |
| Climate and time series score | 47 | 40 |
| Spatial statistics score | 31 | 46 |
| Making maps score | 70 | 82 |
| Recognising infeasible requests score | 63 | 23 |
| Cost per task | $0.021 | $0.10 |
| Time per task (at 100 output tokens/s) | 5.6 min | 12.0 min |
| Tokens per attempt | 168k | 226k |
| Turns per attempt | 14.6 | 13.8 |
| Input price per 1M tokens | $0.10 | $0.70 |
| Output price per 1M tokens | $0.30 | $2.10 |
| Context window | 256,000 | 262,144 |
| Image input | Yes | Yes |
| Developer | Mistral AI |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, Gemma 4 26B A4B (Reasoning), Terminus-2 or Mistral Large 4 (high), Terminus-2?
Mistral Large 4 (high), Terminus-2, 59 to 56.
Which is cheaper per task, Gemma 4 26B A4B (Reasoning), Terminus-2 or Mistral Large 4 (high), Terminus-2?
Gemma 4 26B A4B (Reasoning), Terminus-2, $0.021 to $0.10.
Which writes fewer output tokens per task, Gemma 4 26B A4B (Reasoning), Terminus-2 or Mistral Large 4 (high), Terminus-2?
Mistral Large 4 (high), Terminus-2, 14k to 21k.
Which has the larger context window, Gemma 4 26B A4B (Reasoning), Terminus-2 or Mistral Large 4 (high), Terminus-2?
Mistral Large 4 (high), Terminus-2, 262,144 tokens to 256,000 tokens.