Define which LLM models a workspace can access within an integration, and set a fallback model for when the primary model is unavailable.
| Where Can I Use This? | What Do I Need? |
|---|
- Prisma AIRS AI Gateway (Americas region)
|
- AI Gateway activated
- LLM integration in progress or saved
|
Model provisioning controls the allowlist of models a workspace can call within a given
LLM integration. By default, all models available from the provider are accessible. You can
restrict a workspace to a specific set of models to enforce governance requirements, reduce
cost variance, or prevent the use of deprecated or experimental models. You can also designate
a fallback model that AI Gateway uses automatically when the primary model returns a 429 or
503 from the provider.
