Kimi K2.6 (Reasoning), Kimi Code vs. Qwen 3.8 27B (xhigh), Qwen Code
Kimi K2.6 (Reasoning), Kimi Code against Qwen 3.8 27B (xhigh), Qwen Code: Geospatial Agent Index, each evaluation, cost, time and tokens, on the same tasks and limits. The harnesses differ, so part of any gap can come from the harness. Methodology.
Side by side
| Metric | Kimi K2.6 (Reasoning), Kimi Code | Qwen 3.8 27B (xhigh), Qwen Code |
|---|---|---|
| Geospatial Agent Index: Text only | 69 | 68 |
| Geospatial Agent Index: Text + vision | Not run: Vision agents | Not run: Vision agents |
| Earth-Bench V2 score | 63 | 68 |
| GeoBenchX score | 64 | 62 |
| GS-QA2 score | 80 | 75 |
| 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 |
| Raster and remote sensing score | 75 | 76 |
| Vector and overlay score | 77 | 75 |
| Climate and time series score | 63 | 61 |
| Spatial statistics and interpolation score | 54 | 54 |
| Mapping and cartography score | 74 | 63 |
| Recognising infeasible requests score | 58 | 60 |
| Cost per task | $0.27 | $0.22 |
| Time per task (at 100 output tokens/s) | 3.0 min (estimated) | 4.7 min (estimated) |
| Tokens per attempt | 1.05M | 840k |
| Turns per attempt | No data | 17.6 |
| Attempts scored | 488 | 488 |
| Input price per 1M tokens | $0.95 | $0.45 |
| Output price per 1M tokens | $4.00 | $3.20 |
| Context window | 262,144 | 262,144 |
| Image input | Yes | Yes |
| Developer | Moonshot AI | Alibaba (Qwen) |
Charts
Geospatial Agent Index
Equal average of Earth-Bench V2, GeoBenchX and GS-QA2, 0 to 100 · Higher is better
Top eight. Every model is in the table.
Data table: Geospatial Agent Index
| Model | Developer | Geospatial Agent Index | Status | Evaluations scored | Tasks attempted |
|---|---|---|---|---|---|
| Kimi K2.6 (Reasoning), Kimi Code | Moonshot AI | 69 | Complete | 3 of 3 | 488 of 488 tasks |
| Qwen 3.8 27B (xhigh), Qwen Code | Alibaba (Qwen) | 68 | Complete | 3 of 3 | 488 of 488 tasks |
Cost per task
Average cost per task (USD) · Lower is better
Top eight. Every model is in the table.
Data table: Cost per task
| Model | Developer | Cost per task | Tasks attempted |
|---|---|---|---|
| Qwen 3.8 27B (xhigh), Qwen Code | Alibaba (Qwen) | $0.22 | 488 of 488 tasks |
| Kimi K2.6 (Reasoning), Kimi Code | Moonshot AI | $0.27 | 488 of 488 tasks |
Output tokens per task
Average output tokens per attempt, reasoning included · Lower is better
Top eight. Every model is in the table.
Data table: Output tokens per task
| Model | Developer | Output tokens per task | Tasks attempted |
|---|---|---|---|
| Kimi K2.6 (Reasoning), Kimi Code | Moonshot AI | 18k | 488 of 488 tasks |
| Qwen 3.8 27B (xhigh), Qwen Code | Alibaba (Qwen) | 28k | 488 of 488 tasks |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, Kimi K2.6 (Reasoning), Kimi Code or Qwen 3.8 27B (xhigh), Qwen Code?
Kimi K2.6 (Reasoning), Kimi Code, 69 to 68. Both have attempted all 488 tasks.
Which is cheaper per task, Kimi K2.6 (Reasoning), Kimi Code or Qwen 3.8 27B (xhigh), Qwen Code?
Qwen 3.8 27B (xhigh), Qwen Code, $0.22 to $0.27.
Which writes fewer output tokens per task, Kimi K2.6 (Reasoning), Kimi Code or Qwen 3.8 27B (xhigh), Qwen Code?
Kimi K2.6 (Reasoning), Kimi Code, 18k to 28k.
Which has the larger context window, Kimi K2.6 (Reasoning), Kimi Code or Qwen 3.8 27B (xhigh), Qwen Code?
They are level, at 262,144 tokens.
More: Kimi K2.6 (Reasoning), Kimi Code · Qwen 3.8 27B (xhigh), Qwen Code · All comparisons