The Perplexity Orchestrator API picks the right model for each task and routes to a provider based on availability, capacity and latency. Open-weight and proprietary models sit behind one API with unified billing.