Kimi K2.6 (Reasoning) vs MiniMax M3 (Reasoning)
Results updated 1 November 2026 · In Kimi Code and MiniMax Code, so part of any gap can come from the harness · How we benchmark
Verdict
Kimi K2.6 (Reasoning) edges MiniMax M3 (Reasoning) by 1 point, and is faster; MiniMax M3 (Reasoning) is cheaper.
- Cost per task
- $0.27 vs $0.15
- Time per task
- 3.0 min (estimated) vs 4.9 min (estimated)
- Tokens per attempt
- 1.05M vs 1.68M
- Image input
- Yes vs Yes
Kimi K2.6 (Reasoning), Kimi Code scores higher on GeoBenchX and GS-QA2; MiniMax M3 (Reasoning), MiniMax Code scores higher on Earth-Bench V2. MiniMax M3 (Reasoning), MiniMax Code is cheaper ($0.15 against $0.27 per task), but Kimi K2.6 (Reasoning), Kimi Code is faster (3.0 min against 4.9 min).
Where they differ most
By kind of work, share of tasks passed, 0 to 100 · Higher is better
| Kind of work | Kimi K2.6 (Reasoning) | MiniMax M3 (Reasoning) |
|---|---|---|
| Recognising infeasible requests | 58 | 33 |
| Climate and time series | 63 | 74 |
| Spatial statistics | 54 | 62 |
| Making maps | 74 | 77 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: Kimi K2.6 (Reasoning), Kimi Code · MiniMax M3 (Reasoning), MiniMax Code.
Side by side
| Metric | Kimi K2.6 (Reasoning), Kimi Code | MiniMax M3 (Reasoning), MiniMax Code |
|---|---|---|
| Geospatial Agent Index: Text only | 69 | 68 |
| Geospatial Agent Index: Text + vision | No data | No data |
| Earth-Bench V2 score | 63 | 68 |
| GeoBenchX score | 64 | 59 |
| GS-QA2 score | 80 | 78 |
| Vision agents score | No data | No data |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | No data | 94 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | No data | 79 |
| OGD4All score (standalone; in no index) | No data | 94 |
| GeoNatureAgent score (standalone; in no index) | No data | 93 |
| MultiGlobeQA score (standalone; in no index) | No data | 73 |
| GeoAgentEval score (standalone; in no index) | No data | 92 |
| Remote sensing and raster analysis score | 75 | 73 |
| GIS analysis score | 77 | 81 |
| Climate and time series score | 63 | 74 |
| Spatial statistics score | 54 | 62 |
| Making maps score | 74 | 77 |
| Recognising infeasible requests score | 58 | 33 |
| Cost per task | $0.27 | $0.15 |
| Time per task (at 100 output tokens/s) | 3.0 min (estimated) | 4.9 min (estimated) |
| Tokens per attempt | 1.05M | 1.68M |
| Turns per attempt | No data | No data |
| Input price per 1M tokens | $0.95 | $0.30 |
| Output price per 1M tokens | $4.00 | $1.20 |
| Context window | 262,144 | 1,048,576 |
| Image input | Yes | Yes |
| Developer | Moonshot AI | MiniMax |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, Kimi K2.6 (Reasoning), Kimi Code or MiniMax M3 (Reasoning), MiniMax Code?
Kimi K2.6 (Reasoning), Kimi Code, 69 to 68.
Which is cheaper per task, Kimi K2.6 (Reasoning), Kimi Code or MiniMax M3 (Reasoning), MiniMax Code?
MiniMax M3 (Reasoning), MiniMax Code, $0.15 to $0.27.
Which writes fewer output tokens per task, Kimi K2.6 (Reasoning), Kimi Code or MiniMax M3 (Reasoning), MiniMax Code?
Kimi K2.6 (Reasoning), Kimi Code, 18k to 29k.
Which has the larger context window, Kimi K2.6 (Reasoning), Kimi Code or MiniMax M3 (Reasoning), MiniMax Code?
MiniMax M3 (Reasoning), MiniMax Code, 1,048,576 tokens to 262,144 tokens.