Perplexity has launched the Perplexity Orchestrator API, a single endpoint for frontier and open-weight fashions. It picks a serving provider by availability, capacity and latency, and payments all the things via 1 account. Builders get fewer integrations and 1 bill.
What’s the Perplexity Orchestrator API?
The Perplexity Orchestrator API is a hosted inference gateway that serves many fashions behind 1 base URL. It covers frontier models from OpenAI, Anthropic, Google and xAI. It additionally serves open-weight fashions comparable to GLM, Kimi and DeepSeek, hosted by Perplexity or inference companions.
It accepts OpenAI Chat Completions, OpenAI Responses and Anthropic Messages requests. Most groups solely change the API key and base URL. The OpenAI base URL is https://api.perplexity.ai/orchestrator/v1. Anthropic SDK customers level at https://api.perplexity.ai/orchestrator, with out /v1.
Perplexity runs the identical layer internally, throughout 85+ fashions and greater than a dozen suppliers.
How does supplier failover work?
Failover retains the mannequin fastened and swaps the deployment. A Claude request that fails on Anthropic will be served by AWS Bedrock as an alternative. The reliability docs set these guidelines:
- Triggers are supplier errors, charge limits and timeouts.
- Every try has a time-to-first-token restrict and a total-duration restrict. The values aren’t revealed.
- Invalid or over-context requests return HTTP 400 and aren’t retried.
- If each deployment fails, the API returns HTTP 429 with
Retry-After. - Requests that fail earlier than producing output aren’t billed.
You can’t decide a deployment your self.
What do the reliability and pace numbers present?
Perplexity measured manufacturing visitors from October 4 to November 2, 2026. Claude and GPT requests through Perplexity succeeded more often than requests to any single supplier measured. Day by day success charges additionally various much less throughout supplier incidents. The put up provides no share success charges.
Pace various by supplier too. For Claude Opus 5.5, Anthropic was quickest on 15 of 30 days, AWS Bedrock on 15. Azure’s median output pace for GPT-5.6 Sol was 27% increased than OpenAI’s. Pace-aware routing continues to be in progress.
How does automated routing with perplexity/auto work?
Routing is in preview. You set mannequin to perplexity/auto and move a routing object. It holds a default_model and 1 to 5 routes, every with a reputation, description and mannequin.
A decision model reads the dialog and route descriptions. It by no means sees route names or mannequin IDs. If nothing matches, or the choice mannequin fails or occasions out, the default mannequin solutions. So do media inputs.
Response headers report the outcome: x-perplexity-autoroute-outcome, -route and -confidence. The default coverage sends visitors to GPT-6.1 Sol, with easy requests going to GPT-6 Luna. Perplexity additionally lists built-in Decrease value, Balanced and Increased high quality profiles, plus an Auto effort preview.
(operate(){var f=doc.getElementById(“mtp-pxo-frame”);window.addEventListener(“message”,operate(e){if(e.knowledge&&e.knowledge.mtpPxoH&&e.supply===f.contentWindow){f.fashion.top=e.knowledge.mtpPxoH+”px”;}});})();
Can routing work round refusals?
Perplexity examined this on a 63-task subset of Terminal-Bench 4.1, run in Claude Code. Anthropic documents that Claude Fable 5.2 safeguards can block benign safety and life sciences work.
- Claude Fable 5.2 alone: 38/63 duties.
- Fable plus routes to GPT-6.1 Sol: 41/63.
- Fable plus routes to GPT-6 Astra: 41/63.
The routes coated safety, biology and after_refusal requests. That’s +3 duties, at the next estimated value per process. Perplexity calls it an adjusted comparability; outcomes differ by workload.
How is Orchestrator priced?
Every request is billed on the answering mannequin’s revealed per-token charge. Failover to a different supplier doesn’t change the speed. Perplexity says there’s no additional platform fee on top of credits.
For instance, perplexity/glm-5.3 lists $1.4 input, $4.4 output and $0.26 cache read per 1M tokens. Utilization attracts from pay as you go Perplexity API credit, additionally purchasable via AWS Market.
Entry is tiered by cumulative credit purchases:
- Tier 0: Google, xAI and open-weight fashions.
- Tier 1 ($50): OpenAI and most Anthropic fashions.
- Tier 2 ($250): Claude Opus 5, Opus 5.5 and the Fable fashions.
How does it examine with OpenRouter and Vercel AI Gateway?
| Characteristic | Perplexity Orchestrator | OpenRouter | Vercel AI Gateway |
|---|---|---|---|
| Mannequin catalog | Frontier (OpenAI, Anthropic, Google, xAI) plus open-weight GLM, Kimi, DeepSeek; depend not disclosed (blog) | 400+ fashions, 70+ suppliers (source) | 400+ fashions (source) |
| API codecs | Chat Completions, Responses, Anthropic Messages (docs) | Chat Completions, Responses, Anthropic Messages (docs) | Chat Completions, Responses, Anthropic Messages (docs) |
| Platform charge | No further platform charge on credit (blog) | 5.5% on credit score purchases (Normal), 8% (Enterprise) (pricing) | No platform charge, no token markup (pricing) |
| Token pricing | Answering mannequin’s revealed charge (FAQ) | Supplier worth handed via (FAQ) | Supplier checklist worth (docs) |
| Similar-model supplier failover | Sure, automated (docs) | Sure, automated fallback (source) | Sure, automated (changelog) |
| Immediate-based auto router | perplexity/auto, preview, 1 to five routes (docs) |
openrouter/auto, no router charge (docs) |
Not disclosed |
| Carry your personal key | Not supported (FAQ) | Supported, 5% charge above allowance (FAQ) | Supported, no markup (docs) |
Key Takeaways
- 1 endpoint for OpenAI, Anthropic, Google, xAI and open-weight fashions.
- Similar-model failover throughout suppliers, billed on the similar charge.
perplexity/autoroutes by plain-language descriptions, in preview.- Refusal routing: 38/63 to 41/63 on Terminal-Bench 4.1, self-reported.
Try the launch post, technical blog, quickstart, model catalog and routing docs. You’ll be able to verify the deployment choices and specs right here via our GPULeaderboard.com chrome extension.
Additionally, be at liberty to comply with us on Twitter and don’t overlook to affix our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.
Must accomplice with us for selling your GitHub Repo OR Hugging Face Web page OR Product Launch OR Webinar and so on.? Connect with us
The put up Perplexity Launches Orchestrator API for Failover and Model Routing appeared first on MarkTechPost.

