New: Orchestrator API
By Steven Van ·
Developers can request private preview access and use existing OpenAI or Anthropic SDKs, with billing at each model’s published token rates.
Developers can access open-weight models hosted by Perplexity through one API using their existing Perplexity API key. The Orchestrator API handles routing and failover across model deployments and is currently in private preview.
Use existing OpenAI or Anthropic integrations
The API accepts OpenAI Chat Completions, OpenAI Responses and Anthropic Messages formats. Any model in the catalog works with any supported format, regardless of its original provider. Existing integrations can use the official OpenAI or Anthropic SDK by changing the base URL and supplying a Perplexity API key.
Developers choose a model by its catalog ID and send their own prompts and tools. The OpenAI Responses endpoint is stateless, so each request must include the full conversation.
Automatic routing and per-token billing
The platform routes requests to healthy deployments automatically, with failover handled by Perplexity. Requests are billed per token at the selected model’s published rates, regardless of how they are served, with no per-request fees. The model catalog also lists token prices.
Private preview access
The Orchestrator API documentation directs developers to email api@perplexity.ai to request access.