Gemini 3.1 Pro (preview) (low) vs DeepSeek V4 Flash (high)
Results updated 30 October 2026 · In Terminus-2 and Claude Code, so part of any gap can come from the harness · How we benchmark
Verdict
Not comparable yet: Gemini 3.1 Pro (preview) (low) is not yet ranked.
- Cost per task
- $0.057 vs $0.080
- Time per task
- 0.5 min vs 5.4 min
- Tokens per attempt
- 19k vs 634k
- Image input
- Yes vs No
Gemini 3.1 Pro (preview) (low), Terminus-2 scores higher on GeoBenchX; DeepSeek V4 Flash (high), Claude Code scores higher on Earth-Bench V2. Gemini 3.1 Pro (preview) (low), Terminus-2 is cheaper and faster: $0.057 against $0.080 per task, and 0.5 min against 5.4 min.
Not yet ranked. Gemini 3.1 Pro (preview) (low), Terminus-2: no GS-QA2 score yet. Not in the Agent Index, cost and time charts until it has a score on every index evaluation. Results so far are in the side-by-side table.
Where they differ most
By kind of work, share of tasks passed, 0 to 100 · Higher is better
| Kind of work | Gemini 3.1 Pro (preview) (low) | DeepSeek V4 Flash (high) |
|---|---|---|
| Recognising infeasible requests | 93 | 35 |
| GIS analysis | 80 | 62 |
| Remote sensing and raster analysis | 55 | 67 |
| Making maps | 60 | 68 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: Gemini 3.1 Pro (preview) (low), Terminus-2 · DeepSeek V4 Flash (high), Claude Code.
Side by side
| Metric | Gemini 3.1 Pro (preview) (low), Terminus-2 | DeepSeek V4 Flash (high), Claude Code |
|---|---|---|
| Geospatial Agent Index: Text only | 57 | 56 |
| Geospatial Agent Index: Text + vision | No data | No image input |
| Earth-Bench V2 score | 41 | 50 |
| GeoBenchX score | 74 | 50 |
| GS-QA2 score | No data | 69 |
| Vision agents score | No data | No image input |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 84 | 84 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 79 | 89 |
| OGD4All score (standalone; in no index) | No data | 70 |
| GeoNatureAgent score (standalone; in no index) | No data | 71 |
| MultiGlobeQA score (standalone; in no index) | No data | 64 |
| GeoAgentEval score (standalone; in no index) | No data | 80 |
| Remote sensing and raster analysis score | 55 | 67 |
| GIS analysis score | 80 | 62 |
| Climate and time series score | 45 | 53 |
| Spatial statistics score | 46 | 38 |
| Making maps score | 60 | 68 |
| Recognising infeasible requests score | 93 | 35 |
| Cost per task | $0.057 | $0.080 |
| Time per task (at 100 output tokens/s) | 0.5 min | 5.4 min |
| Tokens per attempt | 19k | 634k |
| Turns per attempt | 3.6 | 16.1 |
| Input price per 1M tokens | $2.00 | $0.44 |
| Output price per 1M tokens | $12.00 | $1.32 |
| Context window | 1,048,576 | 1,048,576 |
| Image input | Yes | No |
| Developer | DeepSeek |