Skip to content
NewA calmer, faster ZeroHour

Perplexity launches Orchestrator API for failover and model routing

Original title: Perplexity Launches Orchestrator API for Failover and Model Routing

MarkTechPostby Michal Sutter ()8 min read1
AI summary · x-ai/grok-4.7Summaries by Vani AI

One endpoint serves frontier and open-weight models, bills the answering model’s rate, and adds no platform fee.

Perplexity’s Orchestrator API is a hosted inference gateway with one account and OpenAI- or Anthropic-compatible requests. Failover keeps the model fixed and switches providers on errors, rate limits and timeouts; failed pre-output requests are not billed. Production measurements from October 4 to November 2 found Claude and GPT traffic through Perplexity succeeded more often than any single measured provider, though no success percentages were published. A preview router, perplexity/auto, can send unmatched or refused work to alternate models.

Unconfirmed: Perplexity calls the refusal-routing comparison adjusted and says results vary by workload.

Key facts
  • OpenAI base URL is https://api.perplexity.ai/orchestrator/v1; Anthropic omits /v1.
  • Invalid or over-context requests return HTTP 400 and are not retried.
  • Total failure returns HTTP 429 with Retry-After; deployments cannot be chosen.
  • Access tiers start at $50 for OpenAI and most Anthropic models, and $250 for Opus and Fable.
  • On a 63-task Terminal-Bench subset, Fable plus GPT routes scored 41/63 versus 38/63 alone.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

8 min read at marktechpost.com

Perplexity’s Orchestrator API serves OpenAI, Anthropic, Google, xAI and open-weight models through 1 endpoint, with same-model provider failover, unified billing and a preview perplexity/auto router that sends each request to a described route. The post Perplexity Launches Orchestrator API for Failover and Model Routing appeared first on MarkTechPost.

MarkTechPost

By Michal Sutter · marktechpost.com

Read the full article at MarkTechPost

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.