Kimi K3

Kimi K3 (moonshotai/kimi-k3) is the Moonshot Kimi flagship: a 2.8T open-weight model for complex coding, knowledge work and long-horizon agentic tasks. It understands images and carries a 1M-token context. Requests go through the usual POST /v1/chat/completions in OpenAI format.

Parameters and capabilities

Capability	Value
model	moonshotai/kimi-k3 (or the short kimi-k3)
Context	1M tokens
Image input	Yes — image_url with a data: URL in the last message
File input	No — a request with a file part returns 400
Video input	Yes — the capabilities.video flag: attachments in the web cabinet and the bot
Reasoning	max only; disabled with off
Cache read	33.40 ₽ ($0.375) per 1M tokens
Cache write	Not published

File input is closed for the model (capabilities.file: false) — unfold documents into text beforehand. There are no intermediate reasoning depths: either the maximum (max) or a full switch-off (off). Reasoning is billed as output tokens and eats into the max_tokens budget.

Examples

A request with maximum reasoning depth

Python

from openai import OpenAI

client = OpenAI(base_url="https://api.mixen.ai/v1", api_key="mxn-...")

resp = client.chat.completions.create(
    model="moonshotai/kimi-k3",
    messages=[{"role": "user", "content": "Design a task queue architecture with delivery guarantees"}],
    reasoning_effort="max",
)
print(resp.choices[0].message.content)

cURL

curl https://api.mixen.ai/v1/chat/completions \
  -H "Authorization: Bearer $MIXEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "moonshotai/kimi-k3",
    "messages": [{"role": "user", "content": "Design a task queue architecture with delivery guarantees"}],
    "reasoning_effort": "max"
  }'

Streaming in Python

stream = client.chat.completions.create(
    model="moonshotai/kimi-k3",
    messages=[{"role": "user", "content": "Break down a contested architectural trade-off and make the call"}],
    stream=True,
    stream_options={"include_usage": True},
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

Price

Per 1M tokens: input — 334.20 ₽ ($3.75), output — 1671.20 ₽ ($18.75). Context cache: read — 33.40 ₽ ($0.375); a separate write price is not published by the catalog. Catalog prices track the exchange rate — take current values from the catalog.

Strengths and limits

• 2.8T open-weight parameters — Moonshot’s declared flagship.
• Complex coding, knowledge work and long-horizon agentic tasks are the model’s core profiles.
• 1M-token context — enough for a whole repository or a large corpus.
• Image input; video attachments — in the web cabinet and the bot.
• No file input: file parts return 400, unfold documents into text in advance.
• No intermediate reasoning depths — max or off; the family has cheaper options: Kimi K2.7 Code for coding (74.60 ₽ input) and Kimi K2.6 for agentic work (105.80 ₽ input).

All models — in the catalog. The general text workflow — chat guide.