Kimi K3

Kimi K3 uses HopBase's OpenAI Chat Completions compatible protocol. Use https://api.hop-base.com/v1 and set the model to kimi-k3.

Group selection and billing

• When creating an API key in the console, set the key's group to Kimi K3 自部署专线 (Kimi K3 Dedicated Line). That key can then call kimi-k3.
• Official list price is $3 per million input tokens and $15 per million output tokens, with cached input reads at $0.30 per million. Your effective rate is shown in the signed-in model catalog.
• One key is bound to one group. Keys bound to other groups get a clear 404 when calling kimi-k3 — create a new key on this group, or rebind an existing key in the console.

Client fields

Reasoning output and prefix caching

• Kimi K3 is a reasoning model. Responses carry a reasoning_content field holding the thinking trace, and reasoning tokens are billed as output tokens.
• Set max_tokens to 1024 or higher. A small budget can be consumed entirely by the thinking trace, leaving a response with reasoning but no final answer.
• Prefix caching is effective in production, matching at a 256-token block granularity. Long multi-turn agent sessions settle at an 88-96% hit rate, so keep the system prompt, tool definitions, and other fixed context at the front of the message list to reuse the cache.

OpenAI SDK / Python

from openai import OpenAI

client = OpenAI(
    base_url="https://api.hop-base.com/v1",
    api_key="sk-your-key",
)

resp = client.chat.completions.create(
    model="kimi-k3",
    max_tokens=4096,
    messages=[{"role": "user", "content": "Explain where Kimi K3 works best"}],
)
message = resp.choices[0].message
print(getattr(message, "reasoning_content", None))
print(message.content)

curl

curl https://api.hop-base.com/v1/chat/completions \
  -H "Authorization: Bearer sk-your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "max_tokens": 4096,
    "messages": [{"role": "user", "content": "Hello, introduce yourself"}]
  }'