A SaaS system serves many similarly structured models with modest individual traffic. Keeping a separate endpoint for each is costly. Which approach should be evaluated?

1. Delete all but one tenant model and return its predictions to every tenant.
2. A supported multi-model endpoint with measured loading behavior and resource isolation needs.
3. Assume every model can share one endpoint regardless of container or hardware compatibility.
4. Keep one maximum-size dedicated endpoint per rarely used model without measurement.