Kimi K3
Moonshot's Kimi — strong agentic reasoning and long context.
Capabilities
Kimi K3 is a language model served through the Pika developer API. Send it a list of chat messages and it streams back a text reply, using the OpenAI-compatible chat-completions shape.
Authentication
api key header
The Pika API uses API keys to authenticate requests. Send your key in the X-API-Key header. Your API key is a secret. Don't expose it in browsers or other client-side code. Instead, call the API from your server. The signed-in playground uses a portal session and never sends your API key.
X-API-Key: YOUR_API_KEY
Endpoint
synchronous · chat completions
Request
POST /v1/chat/completions
Request
curl -X POST https://api.dev.pika.art/v1/chat/completions \-H "X-API-Key: YOUR_API_KEY" \-H "Content-Type: application/json" \-d '{"model": "moonshotai/kimi-k3","messages": [{ "role": "user", "content": "Hello, who are you?" }]}'
Pass moonshotai/kimi-k3 as the model and a messages array of { role, content } turns. Set stream: true to receive tokens incrementally over server-sent events.
The Python and TypeScript examples use the official OpenAI SDK with its base URL pointed at https://api.dev.pika.art/v1 — no Pika-specific SDK is needed. Keep your Pika API key and existing OpenAI code works without a rewrite.
Errors
shared across calls
Errors use conventional HTTP status codes with a JSON body of the shape {"message": "..."}. A missing or invalid key returns 401.
Pricing
billed per token
| Tier | Price |
|---|---|
| Input | $3.00 / 1M tokens |
| Output | $15.00 / 1M tokens |
| Cached input | $0.30 / 1M tokens |
Per-model pricing
Transparent, usage-based rates for every Moonshot AI model.
Kimi K3
LLM