Kimi K3 API
- $2.10 per 1M input tokens
- $10.50 per 1M output tokens
Kimi K3, 30% under list.
We serve Kimi K3, Moonshot AI’s open-weight model, through an OpenAI-compatible API (ARC isn’t affiliated with Moonshot).
What you get
The real model, through an OpenAI-compatible API.
- 01
The real model
It’s the real model: Moonshot’s official K3 checkpoint, unmodified.
- 02
OpenAI-compatible
Use the OpenAI SDK with the base URL we send with your key and the model name kimi-k3.
- 03
Your data
We don’t keep live requests at all, delete batch files 24 hours after a job finishes, don’t train on your data, and process in the United States.
Pricing
Live and batch.
-
Live
30% under K3’s list price.
- Input
- $2.10
- Cached input
- $0.21
- Output
- $10.50
-
Batch
Up to 70% under list, with results within 24 hours.
- Input
- $1.00
- Cached input
- $0.10
- Output
- $4.50
Prices per 1M tokens, in US dollars.
Get started
Email tj@arc.ai for an API key.
We send the base URL with your key. Then call Kimi K3 with the OpenAI SDK:
import os
from openai import OpenAI
client = OpenAI(
base_url=os.environ["ARC_BASE_URL"], # the base URL we send with your key
api_key=os.environ["ARC_API_KEY"],
)
response = client.chat.completions.create(
model="kimi-k3",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)Or copy the address: tj@arc.ai