Kimi K3 API — Access via an AI API Gateway
Call Moonshot AI’s Kimi K3 through byesu — an AI API gateway with pay-as-you-go billing and no subscription. It uses the OpenAI-compatible API, so existing OpenAI SDK code only needs a new base URL and token. Note: Kimi needs a separate token in the Kimi group — see below.
Model specs below are quoted from Moonshot AI’s official documentation (Kimi API Platform), checked on 2026-10-22.
What It Is
Kimi K3 is Moonshot’s most capable flagship model to date, aimed at long-horizon coding, knowledge work and deep reasoning. Moonshot states it has 2.8 trillion parameters in a mixture-of-experts design that activates 16 of 896 experts per token.
| Item | Value |
|---|---|
| Model ID on byesu | kimi-k3 |
| Context window | 1M tokens (1,048,576) |
| Max output | max_completion_tokens defaults to 131,072; input plus output cannot exceed the context window |
| Input → output | Text, images and video → text |
| Thinking | Always on; reasoning_effort accepts low / high / max, default max |
| Tool calling / JSON Schema structured output | Supported |
How Pricing Works
byesu bills per token with no monthly fee. The actual price of every model is shown live in the model list after you sign in, and the console records the charge for each request.
For reference, Moonshot’s official list price is $3.00 input, $0.30 cache-hit input and $15.00 output per million tokens; cache writes are billed by TTL at $3.00 (5 minutes) or $6.00 (1 hour). Moonshot states K3 pricing is not tiered by context length (official pricing page).
Create a token in the Kimi group
kimi-k3 is served by the Kimi group, which is not part of the auto group: a default auto-group token returns “no available channel”. Go to Console → Tokens, create a new token, set its group to Kimi, and call the model with that token. See Choosing a Group.
How to Call It
The base URL depends on the API format:
- OpenAI-compatible (
/v1/chat/completions) →https://byesu.com/v1
Authenticate with Authorization: Bearer sk-YOUR_TOKEN.
Option 1: curl
curl https://byesu.com/v1/chat/completions \
-H "Authorization: Bearer sk-YOUR_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"model": "kimi-k3",
"messages": [
{"role": "user", "content": "Explain what an API gateway is in one sentence"}
]
}'Option 2: Python (OpenAI SDK)
from openai import OpenAI
client = OpenAI(api_key="sk-YOUR_TOKEN", base_url="https://byesu.com/v1")
resp = client.chat.completions.create(
model="kimi-k3",
messages=[{"role": "user", "content": "Explain what an API gateway is in one sentence"}],
)
print(resp.choices[0].message.content)According to Moonshot’s documentation, temperature (fixed at 1.0), top_p (fixed at 0.95), n, presence_penalty and frequency_penalty are fixed for kimi-k3 — leave them out of your requests. Image input must be base64; public image URLs are not supported.
FAQ
How much does the Kimi K3 API cost on byesu?
byesu bills per token with no subscription; the live price is shown in the model list after you sign in. For reference, Moonshot’s official list price is $3.00 input, $0.30 cache hit and $15.00 output per million tokens.
How long is the Kimi K3 context window?
According to Moonshot’s documentation, kimi-k3 has a 1M-token (1,048,576) context window; max_completion_tokens defaults to 131,072, and input plus output cannot exceed the context window.
Why does my default token return “no available channel” for kimi-k3?
Kimi is not part of the auto group. Create a new token in the Kimi group in the console, call kimi-k3 with it, and check that the model ID matches the console exactly.
Why does kimi-k3 return a parameter error?
According to Moonshot’s documentation, temperature, top_p, n, presence_penalty and frequency_penalty are fixed for kimi-k3. Remove them from the request and try again; thinking is always on and cannot be disabled.
Can I call Kimi K3 with the OpenAI SDK?
Yes. Set the SDK base_url to https://byesu.com/v1, use a Kimi-group token as the api_key, and set model to kimi-k3.