DeepSeek V4 Pro (high) vs Qwen 3.8 27B (Reasoning)
Results updated 24 October 2026 · In Terminus-2 and Qwen Code, so part of any gap can come from the harness · How we benchmark
Verdict
DeepSeek V4 Pro (high) edges Qwen 3.8 27B (Reasoning) by 1 point, and is faster; Qwen 3.8 27B (Reasoning) is cheaper.
- Cost per task
- $0.30 vs $0.22
- Time per task
- 4.0 min vs 4.7 min (estimated)
- Tokens per attempt
- 242k vs 840k
- Image input
- No vs Yes
DeepSeek V4 Pro (high), Terminus-2 scores higher on GeoBenchX and GS-QA2; Qwen 3.8 27B (Reasoning), Qwen Code scores higher on Earth-Bench V2. Qwen 3.8 27B (Reasoning), Qwen Code is cheaper ($0.22 against $0.30 per task), but DeepSeek V4 Pro (high), Terminus-2 is faster (4.0 min against 4.7 min).
Where they differ most
By kind of work, share of tasks passed, 0 to 100 · Higher is better
| Kind of work | DeepSeek V4 Pro (high) | Qwen 3.8 27B (Reasoning) |
|---|---|---|
| Spatial statistics | 23 | 54 |
| Making maps | 81 | 65 |
| Climate and time series | 66 | 63 |
| GIS analysis | 73 | 75 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: DeepSeek V4 Pro (high), Terminus-2 · Qwen 3.8 27B (Reasoning), Qwen Code.
Side by side
| Metric | DeepSeek V4 Pro (high), Terminus-2 | Qwen 3.8 27B (Reasoning), Qwen Code |
|---|---|---|
| Geospatial Agent Index: Text only | 70 | 69 |
| Geospatial Agent Index: Text + vision | No image input | No data |
| Earth-Bench V2 score | 68 | 69 |
| GeoBenchX score | 65 | 63 |
| GS-QA2 score | 77 | 75 |
| Vision agents score | No image input | No data |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 96 | 94 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 100 | 79 |
| OGD4All score (standalone; in no index) | No data | 97 |
| GeoNatureAgent score (standalone; in no index) | No data | 100 |
| MultiGlobeQA score (standalone; in no index) | No data | 74 |
| GeoAgentEval score (standalone; in no index) | No data | 88 |
| ImpactMesh score (standalone; in no index) | No data | No data |
| Remote sensing and raster analysis score | 77 | 77 |
| GIS analysis score | 73 | 75 |
| Climate and time series score | 66 | 63 |
| Spatial statistics score | 23 | 54 |
| Making maps score | 81 | 65 |
| Recognising infeasible requests score | 62 | 62 |
| Cost per task | $0.30 | $0.22 |
| Time per task (at 100 output tokens/s) | 4.0 min | 4.7 min (estimated) |
| Tokens per attempt | 242k | 840k |
| Turns per attempt | 13.9 | 17.6 |
| Input price per 1M tokens | $1.32 | $0.45 |
| Output price per 1M tokens | $3.96 | $3.20 |
| Context window | 1,048,576 | 262,144 |
| Image input | No | Yes |
| Developer | DeepSeek | Alibaba (Qwen) |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, DeepSeek V4 Pro (high), Terminus-2 or Qwen 3.8 27B (Reasoning), Qwen Code?
DeepSeek V4 Pro (high), Terminus-2, 70 to 69.
Which is cheaper per task, DeepSeek V4 Pro (high), Terminus-2 or Qwen 3.8 27B (Reasoning), Qwen Code?
Qwen 3.8 27B (Reasoning), Qwen Code, $0.22 to $0.30.
Which writes fewer output tokens per task, DeepSeek V4 Pro (high), Terminus-2 or Qwen 3.8 27B (Reasoning), Qwen Code?
DeepSeek V4 Pro (high), Terminus-2, 20k to 28k.
Which has the larger context window, DeepSeek V4 Pro (high), Terminus-2 or Qwen 3.8 27B (Reasoning), Qwen Code?
DeepSeek V4 Pro (high), Terminus-2, 1,048,576 tokens to 262,144 tokens.