Kimi K3 API pricing: K3, K3.1 Pro and Highspeed per token, with code

The three Kimi K3 models

• Kimi K3 (kimi-k3): Kimi's most capable K3 model, with visual understanding.
• Kimi K3.1 Pro (kimi-k3.1-pro): the newer flagship reasoning model, also with visual understanding, at a lower price than K3.
• Kimi K3.1 Pro Highspeed (kimi-k3.1-pro-highspeed): K3.1 Pro with faster output, at twice the K3.1 Pro price.

All three have a context window of 1,000,000 tokens and take the OpenAI chat completions format. Reasoning tokens are billed as output tokens, so a long chain of thought shows up in the output count.

Kimi K3 API prices (November 2026)

Prices per million tokens. Moonshot's own list prices are from its pricing page, checked on 5 November 2026.

Output is 40% below Moonshot's list as well. OpenRouter lists Kimi K3.1 Pro at $2.00 input and $10.00 output, the same as Moonshot. There is no subscription or minimum: you pay from prepaid credit for the tokens each request used, and a request that gets no answer costs nothing. Credit packs are on the pricing page.

What a request costs

A coding question with 20,000 tokens of context and a 2,000-token answer (reasoning included) costs, on Kimi K3.1 Pro, $0.024 for the input plus $0.012 for the output. The same request on Kimi K3 is $0.036 plus $0.018. Resend the same long prefix and the cached part bills at the cached input rate instead.

Call Kimi K3 with the OpenAI SDK

Point any OpenAI SDK at https://api.revana.si/v1 with your Revana API key and set model to the Kimi id. Streaming works the same way.

import os

from openai import OpenAI

client = OpenAI(base_url="https://api.revana.si/v1", api_key=os.environ["REVANA_API_KEY"])

resp = client.chat.completions.create(
    model="kimi-k3.1-pro",  # or "kimi-k3", "kimi-k3.1-pro-highspeed"
    messages=[
        {"role": "system", "content": "You are a careful senior engineer."},
        {"role": "user", "content": "Find the bug in this function and explain the fix: ..."},
    ],
)
print(resp.choices[0].message.content)
print(resp.usage)  # prompt, cached and completion tokens for this request

curl https://api.revana.si/v1/chat/completions \
  -H "Authorization: Bearer $REVANA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "stream": true,
    "messages": [{"role": "user", "content": "Summarise this contract in five bullet points: ..."}]
  }'

All three models read images: send them as image_url parts in a user message, as with any OpenAI vision request. The docs cover streaming, errors and the usage fields.

Kimi K3 vs Claude Fable 5: price

Kimi K3 is often weighed against Anthropic's Claude Fable 5. We do not publish benchmark scores; this is what each costs per million tokens.

On Revana, Fable 5 is 65% below Anthropic's list price, and both families use the same OpenAI-style request, so you can send one prompt to each and compare the answers and the usage before you choose. The 20,000-in, 2,000-out request above costs $0.07 plus $0.035 on Fable 5.

Routes and limits

Kimi K3 has two routes on Revana, so a request moves to the second when the first is busy or down. Kimi K3.1 Pro and Highspeed have one route each for now. The models page lists every chat, image and video model, and the chat API page lists every chat model with its price.