Orchestrator API
Perplexity launched Orchestrator, an API for proprietary and hosted open-weight models with one bill, same-model provider failover and a preview auto-routing feature.
Orchestrator accepts OpenAI Chat Completions, OpenAI Responses and Anthropic Messages requests, so developers can keep their SDK and change the key, base URL and model ID. If a deployment fails, the request is retried on another deployment of the same model, for example Claude moving from Anthropic to AWS Bedrock. Setting the model to perplexity/auto calls a decision model that picks a model per request, which is a preview feature.
- Date
- Tuesday 17 November 2026
- Lab
- Perplexity
- Kind
- product
- Access
- closed API
- Price
- Pay as you go at each model's published per-token rate
Figures
| Measure | Value | Measured by |
|---|---|---|
| Azure median GPT-5.6 Sol output speed vs OpenAI | 27% higher Oct 4 to Nov 2, 2026 measurement window | company |
The two articles disagree on scope and access. Unite.AI says a Nov 17 launch with Anthropic, OpenAI and Google models, while Crypto Briefing says a private preview from Nov 9 limited to open-weight models.
Sources
- unite.ai/perplexity-launches-orchestrator-api-with-same-model-failover/
- cryptobriefing.com/perplexity-launches-orchestrator-api/
- marktechpost.com/2026/11/19/perplexity-orchestrator-api-failover-model-routing/
This record was partly confirmed: some claims could not be checked on 18 November 2026. How we check