Kimi K2.6 API

Call Kimi K2.6 from the OpenAI SDK for long coding runs and multi-step agent work, from $0.513 per million input tokens.

Get API key Read the docs

262K context
1T total parameters
32B active parameters
FP8 precision served

from openai import OpenAI

client = OpenAI(
    base_url="https://api.modelsheep.com",
    api_key=API_KEY_MODELSHEEP,
)

response = client.chat.completions.create(
    model="moonshotai/Kimi-K2.6",
    messages=[
        {
            "role": "user",
            "content": "Why do sheep count people?"
        }
    ],
)

print(response.choices[0].message.content)

How to use the Kimi K2.6 API

1
Create a ModelSheep account and copy an API key from the dashboard.
2
Point the OpenAI SDK at the base url https://api.modelsheep.com and set the model to moonshotai/Kimi-K2.6.
3
Send your messages to /chat/completions and read the reply from the first choice, but pass chat_template_kwargs with {"thinking": false} in extra_body first if you want an instant answer.

What Kimi K2.6 is

Kimi K2.6 is an open weight mixture of experts model from Moonshot AI. It holds 1T parameters and activates 32B for each token. Moonshot built it for software work that runs long, and reports single runs of more than 4,000 tool calls over more than 12 hours. Against Kimi K2.5 it lifts the agent swarm ceiling from 100 sub-agents and 1,500 steps to 300 sub-agents and 4,000 steps.

What it does well

Handles multi-file code changes in Rust, Go, or Python across long sessions.
Drives sub-agents and thousands of tool calls in one run.
Reads a chart or a screenshot and answers questions about what it shows.

What it is not for

Video input. Moonshot serves video only on its own API. A video URL fails
Short calls requiring fast answers. The model reasons before replying by

Calling it

Post to /chat/completions with the model string moonshotai/Kimi-K2.6. Thinking mode is the default, so the model reasons before it answers. To get an instant answer, send extra_body with chat_template_kwargs set to {"thinking": false}, and lower the temperature, because Moonshot recommends 1.0 for thinking mode and 0.6 for instant mode.

moonshotai/Kimi-K2.6 /chat/completions

SDKs and languages

Kimi K2.6 answers on an HTTP endpoint, so any language that can post JSON reaches it. These are the ones the box above writes out.

OpenAI SDK: PythonTypeScriptGoJavaC#PHPRust
Anthropic SDK: PythonTypeScriptGoJavaC#PHPRust
HTTP: cURLPythonTypeScriptGoRustJavaC#PHP

Specifications

Author	Moonshot AI
Category	language
Accepts	text, image
Returns	text
Total parameters	1T
Active parameters	32B
Score	27.5 on Artificial Analysis intelligence index
Languages	not published
Precision served	FP8
Context	262K tokens
Vector	n/a
Pulls	441K in the last 30 daysWeights pulled from Hugging Face, read on 2026-09-24. It counts the repository, not this API.
License	Modified MIT

Price

Input

$0.513

per million tokens

Output

$2.160

per million tokens

Against OpenRouter

10%

below the cheapest of 9 providers

The cheapest OpenRouter price for this model was $0.570 in and $2.400 out, with a median of $0.950 and $4.000 across the comparable providers. Read on 2026-09-17.

License

Moonshot publishes the weights under a Modified MIT license, which is the MIT text plus one clause. A commercial product above 100 million monthly active users, or above 20 million US dollars in monthly revenue, must show "Kimi K2.6" on its user interface.

Similar models

Minimax M3MiniMaxAI/MiniMax-M3 🤗

Parameters: 427B, 23B
Context: 1M
Input per 1M: $0.252
Output per 1M: $0.990
Released: Jun 2026
Pulls, 30 days: 167.7K
License: MiniMax Community

DeepSeek V4 Prodeepseek-ai/DeepSeek-V4-Pro 🤗

Parameters: 1.6T, 49B
Context: 1M
Input per 1M: $0.939
Output per 1M: $1.879
Released: Apr 2026
Pulls, 30 days: 537.4K
License: MIT

MiMo V2.5XiaomiMiMo/MiMo-V2.5 🤗

Parameters: 310.8B, 15B
Context: 1.1M
Input per 1M: $0.126
Output per 1M: $0.252
Released: Apr 2026
Pulls, 30 days: 282.1K
License: MIT

Start using Kimi K2.6

Point the OpenAI SDK at api.modelsheep.com. Your prompts stay in Germany, and nothing is retained.

Get API key Read the quickstart