Kimi K2.6

by Moonshot AI · served by SonarCompute

Open-weight MoE model with 1T total parameters and 32B active, built for long-horizon agentic work.

Get an API key — $5 free See the code

$5 in free credits on signup · no card required

$0.55 / 1M

Input tokens

$2.20 / 1M

Output tokens

240K

Context window

92 tok/s

Throughput target

FP8

Precision

Model ID — click to copy

// OVERVIEW

What Kimi K2.6 is good at

Kimi K2.6 is Moonshot AI's flagship open-weight release: a mixture-of-experts transformer with a trillion total parameters, of which 32B activate per token. In practice that means dense-model quality at a fraction of the serving cost, and the per-token price passes that saving on.

The model earns its keep on long-horizon agentic tasks: multi-step coding sessions, tool-calling chains, and autonomous workflows that would blow through a smaller context. The 240K window fits an entire mid-size repository plus the conversation that reasons about it, so agents stop losing the thread between steps.

We serve K2.6 in FP8 with streaming enabled by default. Token counts come back in every response, streaming included, so what you meter locally is what shows up on the invoice. Prompts and completions are processed in memory and never written to durable storage.

agentic coding · tool use · repo-scale reasoning · autonomous workflows

// QUICKSTART

One-line switch from OpenAI

# pip install openai
from openai import OpenAI

client = OpenAI(
    base_url="https://api.sonarcompute.com/v1",
    api_key=os.environ["SONARCOMPUTE_KEY"])

client.chat.completions.create(
    model="moonshotai/kimi-k2.6",
    stream=True,
    messages=[{"role": "user",
               "content": "…"}])

Works with the official OpenAI SDKs, LangChain, LiteLLM and any client that accepts a custom base URL. Streaming (SSE) and usage accounting included.

// FAQ

Questions we actually get

Is the Kimi K2.6 API OpenAI-compatible?

Yes. Point the official OpenAI SDK at https://api.sonarcompute.com/v1, use model id moonshotai/kimi-k2.6, and /chat/completions works as-is, including SSE streaming.

How much does Kimi K2.6 cost per million tokens?

$0.55 per 1M input tokens and $2.20 per 1M output tokens, metered exactly — usage accounting is returned in every response.

How do I get a Kimi K2.6 API key?

Create an account at app.sonarcompute.com — new accounts start with $5 in free credits and no card is required.

What context window does Kimi K2.6 support?

240K tokens — enough for a mid-size repository plus the agent conversation on top of it.

Do you log my prompts?

No. Prompts and completions are processed in memory and are not retained after the response is delivered. Details are in our Data Retention & Logging Disclosure.

Also served: MiniMax M2.7

Compare pricing and specs on the models overview, or create an account and test both on the same $5 of free credits.

Create an account MiniMax M2.7 →