Perplexity adds a Multiagent setting with up to 16 collaborating agents

Perplexity added a Multiagent effort level to Computer on November 17, letting a lead agent coordinate up to 16 supporting agents that share findings and check each other's work. The feature is available to Max subscribers on the web and sits above Ultra in Computer's effort controls.
Claude Fable 5.2 leads the default team, with supporting agents using Claude Opus 5.5, GPT 6.1 Sol and Gemini 4 Carbon. Users can request particular models. Perplexity says Fable 5.2 and GPT 6 Astra can also join as supporting agents when requested or needed for a particular subtask.
Agents can exchange evidence directly, question assumptions and coordinate changes to shared files. The lead remains the user's contact, combines the deliverables and passes follow-up changes to affected agents. Stopping the thread stops the team.
Perplexity distinguishes this from Ultra, where supporting agents report only to the lead, and Model Council, where models independently investigate the same question. Multiagent targets interconnected assignments such as diligence reviews and scenario analysis. Its practical change is collaboration between model-backed workers, not simply collecting several independent answers.
Sources and citations
Revision history
Published .