Perplexity Launches Orchestrator API With Same-Model Failover

What happened
Perplexity introduced the Orchestrator API on November 17, 2026. This API allows users to access a range of proprietary and open-weight AI models through a single service. It features failover between different providers for the same model and offers consolidated billing. The system supports models from companies including Anthropic, OpenAI, Google, and others.
Key details
- Provider failover lets a Claude request move from Anthropic to AWS Bedrock if the first provider fails, while still using the Claude model.
- Google, xAI, and open-weight models are available at all usage tiers; OpenAI and Anthropic models require Tier 1 or higher.
- Billing is based on the model that answers the request, using its published per-token rate.
- Between October 4 and November 2, 2026, Orchestrator delivered higher request success rates for Claude and GPT models than any single provider measured.
Why it matters
Developers and companies using AI models now have a way to reduce disruptions from provider outages or slowdowns, as Orchestrator can automatically switch to another provider for the same model. This can help maintain service reliability and control costs, since billing remains consistent regardless of which provider serves the request. Organizations can also simplify their operations with consolidated billing and access to multiple model providers through a single API.
