MiMo V2.6 Flash (Reasoning) vs GPT-5.6 Luna (medium)
Results updated 1 November 2026 · In MiMo Code and Codex, so part of any gap can come from the harness · How we benchmark
Verdict
GPT-5.6 Luna (medium) edges MiMo V2.6 Flash (Reasoning) by 2 points, and is cheaper and faster.
- Cost per task
- $0.047 vs $0.0094
- Time per task
- 5.3 min (estimated) vs 0.8 min
- Tokens per attempt
- 2.16M vs 140k
- Image input
- Yes vs Yes
MiMo V2.6 Flash (Reasoning), MiMo Code scores higher on GS-QA2; GPT-5.6 Luna (medium), Codex scores higher on Earth-Bench V2 and GeoBenchX. GPT-5.6 Luna (medium), Codex is cheaper and faster: $0.0094 against $0.047 per task, and 0.8 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 | MiMo V2.6 Flash (Reasoning) | GPT-5.6 Luna (medium) |
|---|---|---|
| Making maps | 67 | 79 |
| Spatial statistics | 31 | 38 |
| GIS analysis | 83 | 76 |
| Climate and time series | 51 | 54 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: MiMo V2.6 Flash (Reasoning), MiMo Code · GPT-5.6 Luna (medium), Codex.
Side by side
| Metric | MiMo V2.6 Flash (Reasoning), MiMo Code | GPT-5.6 Luna (medium), Codex |
|---|---|---|
| Geospatial Agent Index: Text only | 64 | 66 |
| Geospatial Agent Index: Text + vision | No data | 67 |
| Earth-Bench V2 score | 55 | 56 |
| GeoBenchX score | 57 | 63 |
| GS-QA2 score | 82 | 80 |
| Vision agents score | No data | 69 |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 90 | 98 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 79 | 100 |
| OGD4All score (standalone; in no index) | 84 | No data |
| GeoNatureAgent score (standalone; in no index) | 92 | No data |
| MultiGlobeQA score (standalone; in no index) | 56 | No data |
| GeoAgentEval score (standalone; in no index) | 76 | No data |
| Remote sensing and raster analysis score | 73 | 77 |
| GIS analysis score | 83 | 76 |
| Climate and time series score | 51 | 54 |
| Spatial statistics score | 31 | 38 |
| Making maps score | 67 | 79 |
| Recognising infeasible requests score | 48 | 47 |
| Cost per task | $0.047 | $0.0094 |
| Time per task (at 100 output tokens/s) | 5.3 min (estimated) | 0.8 min |
| Tokens per attempt | 2.16M | 140k |
| Turns per attempt | 23.6 | 9.4 |
| Input price per 1M tokens | $0.14 | $0.20 |
| Output price per 1M tokens | $0.28 | $1.20 |
| Context window | 1,048,576 | 1,050,000 |
| Image input | Yes | Yes |
| Developer | Xiaomi | OpenAI |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, MiMo V2.6 Flash (Reasoning), MiMo Code or GPT-5.6 Luna (medium), Codex?
GPT-5.6 Luna (medium), Codex, 66 to 64.
Which is cheaper per task, MiMo V2.6 Flash (Reasoning), MiMo Code or GPT-5.6 Luna (medium), Codex?
GPT-5.6 Luna (medium), Codex, $0.0094 to $0.047.
Which writes fewer output tokens per task, MiMo V2.6 Flash (Reasoning), MiMo Code or GPT-5.6 Luna (medium), Codex?
GPT-5.6 Luna (medium), Codex, 3k to 32k.
Which has the larger context window, MiMo V2.6 Flash (Reasoning), MiMo Code or GPT-5.6 Luna (medium), Codex?
GPT-5.6 Luna (medium), Codex, 1,050,000 tokens to 1,048,576 tokens.