Deyron Labs graphic: Perplexity Orchestrator, one API for 85+ models, and Multiagent mode with up to 16 agents for Max subscribers; Perplexity blog, 17 November 2026

What happened

Perplexity published two product posts on 17 November 2026: Orchestrator, an API, and Multiagent, a new mode in Perplexity Computer. The descriptions below are Perplexity’s.

Key details

  • Orchestrator: one API over frontier models from OpenAI, Anthropic, Google and xAI plus open-weight GLM, Kimi and DeepSeek; Perplexity says the model layer spans 85+ models and more than a dozen providers. It accepts OpenAI Chat Completions, OpenAI Responses and Anthropic Messages formats and launched through the API Console. Routes and Auto effort are in preview. Perplexity states there is “no additional platform fee on top of credits”; per-token prices are not listed, and volume pricing is by email.
  • Multiagent: an effort level above Ultra in Perplexity Computer, available to Max subscribers on the web. A lead agent (Claude Fable 5.2) splits work among agents running Claude Opus 5.5, GPT 6.1 Sol and Gemini 4 Carbon, which share findings and check each other. The lead can assign up to 16 agents; GPT 6 Astra and Fable 5.2 join as supporting agents when requested or needed. Price is not stated.
  • Not stated: per-model prices in Orchestrator, quality comparisons, and Max pricing for Multiagent.

Why it matters

For developers, a single billing and API layer across providers simplifies switching models, though the post gives no price comparison. Multiagent shows Perplexity using models from competing labs inside one product, including a Claude model as lead agent.

Disclosure: Claude, made by Anthropic, was one of the AI tools used to research and draft this article; the Multiagent feature described uses Anthropic models. Perplexity is not affiliated with Deyron Labs; all claims are Perplexity’s own.

Sources

  1. Introducing Orchestrator: One API for frontier and open-weight models Primary source , Perplexity, 17 November 2026
  2. Introducing Multiagent for the most complex work Primary source , Perplexity, 17 November 2026