Serving
Multi-adapter serving
What is multi-adapter serving?
Multi-adapter serving hosts several fine-tuned adapters against a single resident copy of a base model, switching between them per request. Because adapters are small, the marginal memory cost of an additional variant is a fraction of a full model.
It changes the economics of having many task-specific models. Ten separate merged models need ten times the memory; ten adapters against one base need one base plus ten small files.
The constraint is that every adapter must share the same base model and version. A variant trained against a different base needs its own deployment.
Related terms
Start with the free tier
A magic link creates your account, your tenant and your first API key. No card until you ask for compute.