Moonshot's coding-specialist build of Kimi K2 on Kyma — a trillion-parameter mixture-of-experts model tuned for end-to-end programming, with a 262K-token context window and an always-on thinking mode.
Modalities
Text+Image → Text
Input
$1.009 /1M
Output
$4.774 /1M
Context
262K
Speed
medium
Performance
Live production data from real requests on Kyma — not synthetic benchmarks.
Rank
#25
of 87 active models
Tokens served
2.7M
all-time
Success rate
100%
last 7 days
Top apps using this model
Public apps sending the most traffic to this model — a signal of what real workloads it fits.
Pricing
Pay per token. Cached input is billed at 10% of the input rate.
+ Estimate your workload− Estimate your workload
Estimated monthly cost
$132
$4.40 / day on Kimi K2.7 Code
Same workload on:
Estimates use list pricing. Actual bills depend on real token counts — every response includes its exact cost.
When to use Kimi K2.7 Code
Where this model earns its cost — and where it doesn't.
Kimi K2.7 Code is Moonshot AI's June 2026 coding-focused release in the Kimi K2 family, a 1T-parameter mixture-of-experts model released under a modified MIT license. Where the general K2.6 spreads across agentic, vision, and reasoning work, K2.7 Code is built to complete programming tasks reliably over long contexts, and it always operates in a thinking mode that preserves full reasoning across multi-turn conversations.
Against K2.6 it posts +21.8% on Kimi Code Bench v2 and +31.5% on MLS Bench Lite while using roughly 30% fewer reasoning tokens — so it tends to reach a correct answer with less generated text, which lowers the effective cost on output-heavy coding loops. It accepts both text and image input and supports function calling and structured outputs, so it slots into existing agent frameworks.
On Kyma it routes with automatic failover across multiple serving paths, bills exact per-request cost in usage.cost, and is reachable with the same OpenAI-compatible key as every other model. The 262K-token context window holds a substantial codebase plus the agent's working state, and the always-thinking design suits multi-step tasks over one-shot replies.
End-to-end coding tasks
Its core design target — completing programming work reliably across long contexts rather than answering isolated snippets.
Agentic programming loops
Function calling plus structured outputs keep read-edit-test-iterate agent cycles parseable and on schema.
Cost-sensitive coding
Roughly 30% fewer reasoning tokens than K2.6 means lower output spend on long, output-heavy sessions.
Multimodal code context
Text-plus-image input lets you pass screenshots, diagrams, or UI mockups alongside the code prompt.
Not ideal for: Quick interactive chat where its always-on thinking mode adds latency, or general non-coding agentic work — for that, Kimi K2.6 is the better fit.
How it compares
Against the peers people actually weigh it against.
Quick start
Up and running in under two minutes.
- 1
Create an API key
Sign up and grab a key from the dashboard — $0.50 free credit, no card required.
Get API key → - 2
Make your first request
Drop in your key and send a chat completion — fully OpenAI-compatible.
curl https://kymaapi.com/v1/chat/completions \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "kimi-k2.7-code", "messages": [ {"role": "user", "content": "Explain prompt caching in one paragraph."} ] }' - 3
Stream responses
Add
"stream": trueto receive tokens as they arrive.curl https://kymaapi.com/v1/chat/completions \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "kimi-k2.7-code", "stream": true, "messages": [{"role": "user", "content": "Hello!"}] }'
FAQ
Common questions about this model.