Value coding specialist — 1T MoE agentic coder at budget prices.
Cost-efficient agentic coding
Your agent needs big-repo context (256K cap) or frontier general reasoning.
Compare every model's knowledge cutoff, max output, and context window.
Model ID kimi-k2.7-code; weights on Hugging Face June 12, 2026. Kimi Code membership from $19/mo.
+21.8% over Kimi K2.6 on Kimi Code Bench v2 while using roughly 30% fewer thinking tokens
Only 32B active params per token — fast and cheap to serve at $0.95/$4.00 per 1M (cache hits $0.19)
Modified MIT license with weights on Hugging Face; pairs with the Kimi Code terminal CLI
256K context is a quarter of what 2026 rivals offer for large-repo agent work
Headline gains are on Moonshot's own in-house benchmark; general reasoning lags the Western frontier
What people actually use Kimi K2.7 Code for.
Agentic coding with the Kimi Code terminal CLI at $0.95/1M input
High-volume code review and refactoring where thinking-token burn matters (~30% fewer than K2.6)
Self-hosted coding infra under a modified MIT license
The nearest models people weigh against it, and what actually separates them.
vs Codestral 25.01 — Against Codestral 25.01 (Mistral), Kimi K2.7 Code costs about 27% more per token and answers slower. Codestral 25.01 is the one to check first if the price difference matters more than the ceiling.
vs DeepSeek V4-Flash — Against DeepSeek V4-Flash (DeepSeek), Kimi K2.7 Code costs about 92% more per token and gives up 3.9x on context. DeepSeek V4-Flash is the one to check first if the price difference matters more than the ceiling.
vs DeepSeek V4-Pro — Against DeepSeek V4-Pro (DeepSeek), Kimi K2.7 Code costs about 74% more per token, gives up 3.9x on context and answers faster. DeepSeek V4-Pro is the one to check first if the price difference matters more than the ceiling.
Price History
→0% since Aug 7
38 data points · tracked daily since Aug 7, 2026
Cost-efficient agentic coding. Start free — no card required.
Recommendations are made independently based on real-world use and public benchmarks. See our disclosures for details.
Similar models worth checking before you commit.
Coding-specialist model designed for fast engineering assistance at a budget-conscious price point.
A 284B-parameter (13B active) MoE workhorse re-post-trained for agentic and coding tasks — beats the V4-Pro preview on every published agent benchmark at ultra-commodity pricing.
DeepSeek's 1.6T-parameter (49B active) MoE flagship with hybrid sparse attention — near-frontier coding and reasoning at roughly a tenth of closed-rival pricing, MIT-licensed open weights.
Kimi K2.7 Code costs $0.95 per million input tokens and $4 per million output tokens on the API, with cached input at $0.16 per million. A month of 10M input and 2M output tokens runs about $17.50 at list price, before any batch or caching discounts.
Kimi K2.7 Code has a 256k tokens context window, with up to 33k tokens of output per response. That is the total of prompt plus response the model can hold in one request.
Kimi K2.7 Code's training data runs through January 2025, and the model was released on June 12, 2026. For anything after that date it needs web search or documents in the prompt.
Kimi K2.7 Code is best for cost-efficient agentic coding. It is a strong fit when that workflow matters more than the tradeoffs around budget pricing and fast speed.
Your agent needs big-repo context (256K cap) or frontier general reasoning.
DeepSeek V4-Pro (DeepSeek) at $0.43/1M/1M input against Kimi K2.7 Code's $0.95/1M/1M — roughly 74% less per token all in. Best open-weights flagship — near-frontier coding at a tenth of the price. Compare it first if Kimi K2.7 Code's pricing is the thing stopping you.
Codestral 25.01 — very fast against Kimi K2.7 Code's fast, with 256k tokens of context. Worth the swap when response time is what your users notice rather than the last few points of reasoning depth.
Newsletter
We track pricing daily. When this model drops or spikes, you'll know first.
No spam. Useful updates only. Affiliate disclosures always clearly labeled.
No reviews yet — be the first.