Perplexity Orchestrator API Guide: One Key for Claude, GPT, Gemini, Grok and Open-Weight Models

Sami Ullah Khan

November 18, 2026

Perplexity Orchestrator API Guide: One Key for Claude, GPT, Gemini, Grok and Open-Weight Models

Perplexity’s Orchestrator API gives developers one API key and one endpoint for a catalogue of models from Anthropic, OpenAI, Google, xAI and open-weight labs, billed on a single Perplexity invoice. It is in private preview: Perplexity’s documentation asks developers to email api@perplexity.ai to request access. This guide covers how it works, what it costs and when it is worth using.

Where it came from

The product has changed name twice. Perplexity launched a Gateway API in August 2026 for open-weight models it hosts itself, renamed it the Router API later that month and moved it behind a private-preview gate, then rebranded it as the Orchestrator API with a new endpoint, according to pricing tracker UsagePricing. Crypto Briefing dates the current private preview to 9 November 2026.

How to connect

The API accepts three request formats with the same prefixed model IDs: OpenAI Chat Completions, OpenAI Responses and Anthropic Messages. That means most existing code needs only a new base URL and key.

SDKBase URL
OpenAI SDKhttps://api.perplexity.ai/orchestrator/v1
Anthropic SDKhttps://api.perplexity.ai/orchestrator (no /v1, because the SDK appends /v1/messages)

A request to /orchestrator/v1/models returns the models your account tier can call, with base rates.

Sample prices (per 1M tokens)

Model IDInputOutput
anthropic/claude-fable-5-2$8.00$40.00
anthropic/claude-sonnet-5-5$2.00$10.00
openai/gpt-6.1-sol$2.00$10.00
openai/gpt-6-luna$0.10$0.50
google/gemini-3.8-flash$0.75$3.75
xai/grok-4.7$2.00$6.00
perplexity/glm-5.3-flash$0.15$0.50
perplexity/deepseek-v4-pro-0813$1.32$3.96
perplexity/kimi-k3$3.00$15.00

Prices are from Perplexity’s model catalogue at the time of writing; some OpenAI and xAI models charge more for very long inputs. Pricing is per token at each model’s published rate, with no per-request fee. The “perplexity/” prefix shows who serves the model, not who created it: GLM, DeepSeek and Kimi are open-weight models that Perplexity hosts on its own infrastructure.

Billing rules worth knowing

  • Reasoning tokens are billed at the output rate.
  • Cache writes for Anthropic and recent OpenAI models cost 1.25 times the input rate; Anthropic one-hour cache writes cost twice the input rate.
  • Usage tiers: calling a model above your organisation’s tier returns HTTP 402.
  • Zero data retention: Claude Fable models are unavailable to organisations with zero data retention, and those requests return HTTP 403.
  • Automatic routing: the perplexity/auto option picks a model per request, and you pay that model’s rate.

When to use it

The Orchestrator API suits teams that want to test several providers without separate contracts, or route cheap tasks to open-weight models and hard tasks to frontier models through one integration. If you only ever use one provider, a direct contract may give you newer models first: at the time of writing, the catalogue lists GPT-6.1 rather than the newly released GPT-6.2.

FAQ

Is the Perplexity Orchestrator API free?

No. You pay per token at each model’s rate. Access is currently by request during the private preview.

Does it include Perplexity’s search?

The Orchestrator API is for model access. Perplexity sells web search separately through its Search API.

Sources

Stay Ahead of AI

Get the latest AI news delivered to your inbox.

We don’t spam! Read our privacy policy for more info.

Published

Spotted an error? Read our corrections policy or email editorial@perplexityaimagazine.com.