DeepSeek V4 Pro (high) vs MiMo V2.6 Flash (Reasoning)
Results updated 1 November 2026 · In Terminus-2 and MiMo Code, so part of any gap can come from the harness · How we benchmark
Verdict
MiMo V2.6 Flash (Reasoning) edges DeepSeek V4 Pro (high) by 1 point, and is cheaper; DeepSeek V4 Pro (high) is faster.
- Cost per task
- $0.30 vs $0.047
- Time per task
- 4.0 min vs 5.3 min (estimated)
- Tokens per attempt
- 242k vs 2.16M
- Image input
- No vs Yes
DeepSeek V4 Pro (high), Terminus-2 scores higher on GeoBenchX; MiMo V2.6 Flash (Reasoning), MiMo Code scores higher on GS-QA2; they are level on Earth-Bench V2. MiMo V2.6 Flash (Reasoning), MiMo Code is cheaper ($0.047 against $0.30 per task), but DeepSeek V4 Pro (high), Terminus-2 is faster (4.0 min against 5.3 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) | MiMo V2.6 Flash (Reasoning) |
|---|---|---|
| Spatial statistics | 23 | 38 |
| GIS analysis | 73 | 83 |
| Making maps | 81 | 72 |
| Climate and time series | 66 | 63 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: DeepSeek V4 Pro (high), Terminus-2 · MiMo V2.6 Flash (Reasoning), MiMo Code.
Side by side
| Metric | DeepSeek V4 Pro (high), Terminus-2 | MiMo V2.6 Flash (Reasoning), MiMo Code |
|---|---|---|
| Geospatial Agent Index: Text only | 70 | 71 |
| Geospatial Agent Index: Text + vision | No image input | No data |
| Earth-Bench V2 score | 68 | 68 |
| GeoBenchX score | 65 | 64 |
| GS-QA2 score | 77 | 82 |
| Vision agents score | No image input | No data |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 96 | 92 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 100 | 84 |
| OGD4All score (standalone; in no index) | No data | 91 |
| GeoNatureAgent score (standalone; in no index) | No data | 100 |
| MultiGlobeQA score (standalone; in no index) | No data | 68 |
| GeoAgentEval score (standalone; in no index) | No data | 96 |
| ImpactMesh score (standalone; in no index) | No data | No data |
| Remote sensing and raster analysis score | 77 | 78 |
| GIS analysis score | 73 | 83 |
| Climate and time series score | 66 | 63 |
| Spatial statistics score | 23 | 38 |
| Making maps score | 81 | 72 |
| Recognising infeasible requests score | 62 | 60 |
| Cost per task | $0.30 | $0.047 |
| Time per task (at 100 output tokens/s) | 4.0 min | 5.3 min (estimated) |
| Tokens per attempt | 242k | 2.16M |
| Turns per attempt | 13.9 | 23.6 |
| Input price per 1M tokens | $1.32 | $0.14 |
| Output price per 1M tokens | $3.96 | $0.28 |
| Context window | 1,048,576 | 1,048,576 |
| Image input | No | Yes |
| Developer | DeepSeek | Xiaomi |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, DeepSeek V4 Pro (high), Terminus-2 or MiMo V2.6 Flash (Reasoning), MiMo Code?
MiMo V2.6 Flash (Reasoning), MiMo Code, 71 to 70.
Which is cheaper per task, DeepSeek V4 Pro (high), Terminus-2 or MiMo V2.6 Flash (Reasoning), MiMo Code?
MiMo V2.6 Flash (Reasoning), MiMo Code, $0.047 to $0.30.
Which writes fewer output tokens per task, DeepSeek V4 Pro (high), Terminus-2 or MiMo V2.6 Flash (Reasoning), MiMo Code?
DeepSeek V4 Pro (high), Terminus-2, 20k to 32k.
Which has the larger context window, DeepSeek V4 Pro (high), Terminus-2 or MiMo V2.6 Flash (Reasoning), MiMo Code?
They are level, at 1,048,576 tokens.