Start for $5

The catalog

Every model, one key

Pin a model by name and get exactly that model, or leave it out and let the router choose.

  • 1000+Models
  • 50+Labs
  • 4Modalities
  • 1API key

Direct access

Pinned by name, never substituted

Pin a name and you get that model or nothing: no substitution, you pay its price plus Corent’s fee, and the receipt names the model you pinned. Quality is Corent’s score out of ten, measured from independent benchmarks and our own test runs, and updated as new results come in.

1000+ models

Image

POST /v1/images/generate
LabPin by nameQuality

Video

POST /v1/videos/generate
LabPin by nameQuality

Voice

POST /v1/audio/speech
LabPin by nameQuality

Text

POST /v1/chat/completions
LabPin by nameQuality

Two ways to call

Name a model, or let Corent choose

You pick

POST /v1/images/generate{  "prompt": "a lighthouse at dusk",  "model": "corent-flux-schnell"}

Direct access runs that model and no other. A video still costs more for a longer clip or a higher resolution, set to the nearest length and resolution that model offers.

The router picks

POST /v1/images/generate{  "prompt": "a lighthouse at dusk",  "quality": "best"}

Leave quality out and Corent picks the cheapest model that does the job well. Add "quality": "best" for the top one. Either way the router reads the prompt and switches models if one fails.

On the OpenAI-compatible chat endpoint, "corent" in the model field lets the router choose. For images, video and voice you can also name a model in the prompt: “use seedance 2.5: a fox running in snow”.

What each kind of result costs, and how to call it

PricingDocs