Direct
Official APIs
A direct commercial and technical relationship with the model company or its official platform.
Your app → Official API → Model runtime
Provider discovery
Chinese AI providers and inference routes
The same model name can sit behind very different companies, infrastructure and data paths. Start by identifying who serves the model and how each request travels.
Three common paths
Separate the model from the route.
Routed
Aggregators and providers
A gateway or inference provider sits between your application and the model runtime.
Your app → Gateway → Inference provider → Model
Private
Self-hosted infrastructure
Your team operates the compatible model weights in local, VPC or on-premises infrastructure.
Your app → Private runtime → Model weights
The provider changes the data path.
Provider statements, observed infrastructure and contractual commitments are different kinds of evidence. Unknown information should remain unknown.