Make your first request.

Create an API key in your console, then use your deployment’s origin with the /v1 base path.

curl https://YOUR_DOMAIN/v1/chat/completions \
  -H 'Authorization: Bearer YOUR_BLACK_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "zhipu/glm-5.3-derisked",
    "messages": [{"role": "user", "content": "Hello"}],
    "max_tokens": 512
  }'

API endpoints

GET /v1/models returns the configured catalogue and prices. POST /v1/chat/completions returns a non-streaming chat completion. Both require your black.ai bearer key.

Text messages support system, user, and assistant roles. Optional parameters: max_tokens (1–8192), temperature (0–2), and stream: false. Tools, images, and streaming are not enabled in this version.

Wallet billing

Each request reserves the model’s full input and output context price, then releases the unused portion after usage is returned. This conservative hold is $4.50 for GLM 5.3 De-risked. Only the actual reported tokens are charged. Timeouts or missing usage leave a hold for operator review to prevent duplicate charges or unaccounted spending.

Insufficient funds return 402; invalid requests return 400; invalid keys return 401; rate limits return 429; provider failures return 502. Never automatically retry an uncertain request.

Access and privacy

The site uses one essential session cookie. There are no browser scripts, external fonts, or analytics trackers. Login codes and API keys are stored as hashes. API prompt content is not persisted. Chat conversations are stored for recovery: temporary chats expire after 24 hours; normal chats remain until deleted. Account, request, and billing metadata are retained for usage and cost accounting.

Tor hides your network address from this service when using its onion address. It does not change the privacy policies of services used to process a request. Service policies still apply.