Quick Start
⏳ File retention:Files generated or uploaded on the platform (images, videos, audio and other outputs) are kept for 15 days and then deleted automatically; their links stop working once deleted. Please download them and re-host them in your own storage in time.
CQT AI offers AI generation across images, video, music, and chat. You can start calling the APIs in three steps: register & top up → create an APIKey → send your first request.
Onboarding flow
- Register and top up — After creating an account, head to the How to Top Up page to purchase credits.
- Create an APIKey — Create a key in APIKey Management, in the form
sk-cqt-xxxx. - Pick a model and call it — Choose an image / video / music / chat model from the left sidebar by use case, then copy the sample code and swap in your APIKey.
🔑 Don't have a key yet? Go straight to APIKey Management to create one.
Common conventions
- Submit a task:
POST /api/cqt/generator/{model} - Query the result:
GET /api/cqt/info/{model}?id={taskId} - Auth header:
Authorization: Bearer {API_KEY} - Responses are uniformly
{ "code":200, "msg":"success", "data": ... }; on a successful submission,datais the taskId.
Your first request
curl -X POST https://api.cqtai.com/api/cqt/generator/nano \
-H "Authorization: Bearer sk-cqt-xxxx" \
-H "Content-Type: application/json" \
-d '{ "model": "nano-banana", "prompt": "a cute cat" }'
# → { "code":200, "data":"<taskId>" }Async tasks return a taskId. Poll for the result by taskId until status = succeeded / failed.
List available models and prices
GET /api/cqt/models returns every model currently on offer together with its price. No authentication is required. The response follows the shape of OpenAI's GET /v1/models, so the OpenAI SDK's models.list() can read it directly.
curl https://api.cqtai.com/api/cqt/models
# Filter by type: ?category=chat | image | video | audio{
"object": "list",
"data": [
{
"id": "gpt-4o",
"object": "model",
"created": 0,
"owned_by": "openai",
"category": "chat",
"name": "gpt-4o",
"pricing": {
"currency": "credits",
"unit": "token",
"input_per_million": 2500,
"output_per_million": 10000,
"cache_write_per_million": 3125,
"cache_read_per_million": 250,
"rate": 1
}
},
{
"id": "happyhorse-1.0-i2v",
"object": "model",
"created": 0,
"owned_by": "alibaba",
"category": "video",
"channel": "qwen",
"pricing": {
"currency": "credits",
"unit": "per_second",
"tiers": [
{ "key": "720P", "label": "720P", "unit": "per_second", "price": 90 },
{ "key": "1080P", "label": "1080P", "unit": "per_second", "price": 160 }
]
}
}
]
}The models and numbers above are examples only. The live response is authoritative.
Fields:
| Field | Description |
|---|---|
id | Model name — the model value you pass when calling |
owned_by | Model vendor |
category | Type: chat / image / video / audio, etc. |
channel | Generation models only; submit via POST /api/cqt/generator/{channel} |
created | Kept for OpenAI compatibility; always 0 |
pricing.currency | Always credits (100 credits = 1 CNY) |
pricing.unit | Billing unit: token / per_call / per_image / per_second / per_task |
Pricing notes:
- Chat models (
unit: token):input_per_million/output_per_millionare credits per million tokens; the actual charge = usage × unit price ×rate.cache_*fields only exist for OpenAI / Claude models. Flat-price chat models (unit: per_call) only havepriceand are not multiplied byrate. - Generation models: a single price is given as
price; models with several tiers (resolution, duration, …) leavepriceempty and list each tier intiers.tiersis a price list — whether tiers are mutually exclusive or additive depends on the model, so do not simply add them up; see the model's own API page for how they combine. - A model without
price/tiersdoes not necessarily have an unpublished price: it may be a free preparatory step (price 0), a post-paid model whose final price can't be known at submission, or one that isn't published yet. When in doubt, check the Pricing page or get a quote for your actual parameters. - Generation models also carry
endpoint_mode:asyncmeans submit viaPOST /api/cqt/generator/{channel};syncmeans a direct synchronous endpoint (e.g.minimax-h3usesPOST /v1/videos). - If your account has a personal discount, the actual charge is further multiplied by it. To see prices with your discount applied, or to get a quote for a specific request before submitting, see Model Pricing Lookup.