Model orchestration manages how an application selects and combines models. It can route requests by task, latency, cost, modality, availability or quality and can add fallbacks when a preferred provider is unavailable.
Continuous evaluation across models and providers supports evidence-based routing. An orchestration layer should preserve consistent grounding and validation while models change, because a faster model is not useful if its outputs fail the site's accuracy or brand requirements.
Acronyms and aliases
AI model orchestration variant
Specialised terms
Related terms
Frequently asked questions
Why use more than one AI model in an application?
Models differ in speed, cost, context, modality and quality. Routing each task to an appropriate model can improve reliability and operating economics.
What does a model orchestrator need to monitor?
It should monitor availability, latency, cost, output quality, safety and task success, then apply clear fallback and validation rules when conditions change.
