Kimi K3 API Guide: Pricing, Context and Code

Quick start

Use the official Kimi endpoint and keep the API key in an environment variable. The launch documentation identifies `kimi-k3` as the general API model name; Kimi Code uses the shorter `k3` identifier in its own clients.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["MOONSHOT_API_KEY"],
    base_url="https://api.moonshot.cn/v1",
)

response = client.chat.completions.create(
    model="kimi-k3",
    messages=[{"role": "user", "content": "Review this repository"}],
    reasoning_effort="max",
)

Launch pricing

Kimi's launch post lists prices per one million tokens. Cached input is much cheaper than uncached input, so stable prompts and reusable context can materially reduce cost. Treat these as launch figures and confirm the live platform price before making a budget commitment.

Context and reasoning choices

K3 supports a context window of up to one million tokens, but a larger window is not a reason to send everything. Retrieve the most relevant files, preserve the model's thinking history when the client supports it, and start a fresh session when changing models or reasoning effort.

• Use max effort for difficult coding, research and multi-step agent work.
• Keep system instructions explicit because K3 can be overly proactive.
• Measure cache hit rate, input tokens, output tokens and task success together.

Frequently asked questions