Perplexity Orchestrator API: 85+ Models, One Key, Same-Model Failover Perplexity launched its Orchestrator API on November 17, a single base URL that routes requests across 85+ models from Anthropic, OpenAI, Google, xAI and others on one bill. The API's distinguishing feature is same-model failover: when a provider goes down, Orchestrator retries a Claude request on a different Claude deployment rather than silently switching the caller to a different model such as GPT. Perplexity launched its Orchestrator API on November 17 — a single base URL that routes requests across 85+ models from Anthropic, OpenAI, Google, xAI, and others on one bill. The headline model count is not the story. The story is same-model failover: when a provider goes down, Orchestrator retries your Claude request on a different Claude deployment instead of quietly swapping you to GPT. That distinction matters a lot in production. Why Same-Model Failover Changes the Reliability Equation Every LLM router on the market today handles provider failures the same way: it falls back to a different model. If Anthropic … The post