Documentation
How to use InferX
Three paths to running your model. Pick based on what you need — you can always switch later.
Whichever path you choose — your traffic runs on a dedicated serverless instance with an OpenAI-compatible API, ready for OpenCode, Dify, OpenWebUI, and more.
Choose your path
EndpointsFastest path to a working API
No deployment needed. Call a ready-made endpoint and get an API key instantly.
- → No snapshot retention fee
- → No deployment to manage
- → Grab an API key and go
- → Traffic on your own instance
CatalogKnown model, some customization
Start from a curated template, deploy it into your own tenant.
- → Model in catalog but not Endpoints
- → Safe curated starting point
- → Customize before deploying
- → Runs in your own tenant
Deploy from ScratchFull control over runtime & config
Define the model spec, runtime, and environment yourself.
- → Model not in Catalog
- → Custom env variables
- → Custom runtime behavior
- → Own the full config
Endpoints — when to use it
Best when you want the fastest integration path and don't need to deploy or manage a model yourself. No snapshot retention fee is charged, and your traffic still runs on your own dedicated instance.
Open Endpoints →Catalog — when to use it
Best when Endpoints doesn't have the model you need, or you want to customize from a curated starting point before deploying into your own tenant. Safer and faster than starting from scratch.
Open Catalog →Deploy from Scratch — when to use it
Best when neither Endpoints nor Catalog covers your model or runtime setup. Full control over model choice, runtime commands, environment variables, and deployment shape.
Create a model →Rule of thumb
- 1Start with Endpoints when you want the fastest integration and don't need to deploy a model yourself.
- 2Move to Catalog when Endpoints doesn't have what you need, or you want a safe customizable starting point.
- 3Use Deploy from Scratch only when Catalog still doesn't cover the model or runtime setup you need.