Kimi K2.7 Code
Moonshot AI released 2026-06-12 modified-mit
MiniMax M3 is both better and cheaper.
It scores +2.4 higher and costs 61% less ($0.525/M against $1.35/M) on this workload — and it does everything this model does.
DeepSeek V4 Flash 0731 is cheaper still (92% less) but drops no image input.
Our take
editorial — not a measurementA coding-specialised open-weight model at $0.67/$3.40 with a 262k window, and the sharpest price contrast in the dataset: it is roughly a quarter the price of GPT-5.3-Codex ($1.75/$14) on output, the dimension that dominates codegen cost. If you are paying for a coding model and have not benchmarked this one against your actual repo tasks, you are probably overspending. Open weights mean you can also self-host it behind your firewall, which matters for proprietary codebases.
Strengths
- $0.67/$3.40 — ~4x cheaper output than GPT-5.3-Codex
- 262k context and 262k max output
- Open weights — self-hostable for proprietary code
- Cache reads at $0.17
Weaknesses
- No independent benchmark score captured in this pass
- Licence not independently verified
- Smaller context than the 1M-token generalists
Reach for it when
- High-volume code generation
- On-prem coding assistants over proprietary repos
- Cost-reduction target for teams on Codex
Avoid it if
- You need verified benchmark evidence before adopting
- Your tasks need >262k of repo context
Sources: openrouter.ai
Every price dimension
| Input | $0.670/M |
|---|---|
| Output | $3.40/M |
| Cached input | $0.170/M |
Independent scores
| Intelligence | 43 |
|---|---|
| Coding | 60.8 |
| Agentic | 30.3 |
Sources are listed separately rather than averaged. Across the 62 models both have scored they correlate at r = 0.806 — close agreement overall, but four models rank very differently between them, and a blended score would hide exactly those.
Best rankings by task
- htmlslides #7of 34 1215
- godotgamedev #12of 43 1183
- asciiart #14of 83 1243
- python-pptxslides #17of 35 1168
- agenticgamedev #17of 34 1138
- website #18of 151 1292
- fullstack #19of 59 1208
- webapps #21of 56 1212
Capability
| Context window | 262K |
|---|---|
| Max output | 262K |
| Input modes | text, image |
| Tool use | yes |
| Reasoning | always on |
| Open weights | yes |
Provenance
| Price source | openrouter.ai |
|---|---|
| Fetched | 2026-08-24 |
| Quality data | verified |
| Cross-checked | vendor page |
MoonshotAI: Kimi K2.7 Code is a coding-focused model in Moonshot AI's Kimi K2 family, built to complete end-to-end programming tasks reliably over long contexts. It uses a native multimodal mixture-of-experts...