Kimi K2.7 Code

Moonshot AI's open-weight coding model (June 2026): a 1T-parameter MoE (32B active) that uses ~30% fewer reasoning tokens than K2.6 and posts strong tool-use scores. Input cache-hit as low as $0.19/1M; third-party hosts (OpenRouter/DeepInfra) ~$0.74 in / $3.50 out. Modified-MIT open weights, deployable via vLLM/SGLang. Benchmarks are vendor-reported — verify.

Kimi K2.7 Code strengths

• Open weights (Modified MIT)
• Repository-scale agentic coding
• ~30% fewer reasoning tokens vs K2.6
• Strong tool-use (MCP)
• Self-hostable (vLLM/SGLang)

Pricing & context

Cost guide: a typical call of ~10K input + 2K output tokens runs roughly $0.95 × 0.01 + $4.00 × 0.002 — worth modelling against cheaper tiers before committing high-volume traffic.

When to choose Kimi K2.7 Code

Kimi K2.7 Code is best for Repository-scale refactoring and long multi-turn agentic coding with an open-weight, cost-efficient model. If your workload is more cost-sensitive, weigh it against Llama 4 Scout (~$0.08 (varies by host) input /1M) first.

Kimi K2.7 Code FAQ

How much does Kimi K2.7 Code cost?

Kimi K2.7 Code is priced at $0.95 per 1M input tokens and $4.00 per 1M output tokens (public API list price), with a 256K tokens context window.

What is Kimi K2.7 Code best for?

Kimi K2.7 Code by Moonshot AI is best for Repository-scale refactoring and long multi-turn agentic coding with an open-weight, cost-efficient model.

Is Kimi K2.7 Code multimodal?

Kimi K2.7 Code supports text, image, video.