A Guth Labs publication ↗
Back to News

Perplexity launches Orchestrator API with shared billing and model failover

Metal ring with glass pathways and a backup loop.
AI-generated editorial illustration

Perplexity launched its Orchestrator API on November 17, combining access to proprietary and open-weight models with consolidated billing and provider failover. Its announcement lists support for OpenAI, Anthropic, Google and xAI models alongside families such as GLM, Kimi and DeepSeek.

The service accepts OpenAI Chat Completions, OpenAI Responses and Anthropic Messages formats, allowing applications to retain familiar SDKs. Perplexity says there is no additional platform fee on top of credits. Console controls include spending limits, model allowlists and usage breakdowns by model, project, key and date.

Provider failover keeps the selected model rather than substituting another one. According to the technical explanation, errors, rate limits or timeouts can trigger another attempt on a different deployment of that model.

Separate Routes controls let teams describe which requests should use particular models and supply a default. Auto effort, in preview, adjusts a supported model's reasoning effort according to a developer policy.

The distinction matters for applications balancing model choice with reliability: routing between models and recovering from a provider outage are different decisions, now exposed through the same service.

Sources and citations

Revision history

Published .