Perplexity Launches Multiagent to Coordinate Frontier AI Models

Unite.AI
Read full story· 4 min
Perplexity launched Multiagent, a new coordination mode for its Computer agent platform that assigns up to 16 supporting agents across multiple frontier AI models to collaboratively analyze complex tasks, cross-check findings, and deliver unified responses. The system enables direct communication among agents running Claude, GPT, and Gemini models, enhancing accuracy and consistency in multi-part assignments like financial reviews and due diligence. Multiagent is available to Perplexity Max web订

What happened

On November 17, 2026, Perplexity introduced Multiagent, a new mode for its Computer agent platform. This feature allows a lead agent to assign up to 16 supporting agents, each running on different advanced AI models, to work together on complex tasks. The agents can directly communicate, share evidence, and cross-check their findings before producing a single, unified response. Multiagent is now available to all Perplexity Max subscribers using the web platform.

Key details

  • Multiagent supports up to 16 agents per assignment.
  • The default team uses Claude Fable 5.2 as lead, with Claude Opus 5.5, GPT 6.1 Sol, and Gemini 4 Carbon as supporting models.
  • Users can select which models to include and monitor agent activity during a task.
  • Agents can coordinate edits to shared files and notify each other of changes.
  • Multiagent is accessible through the effort slider for Perplexity Max web subscribers.

Why it matters

Perplexity Max subscribers can now assign multiple AI models to work together on tasks that require detailed analysis and cross-checking, such as financial reviews or due diligence. This coordination reduces the risk of inconsistent findings and streamlines workflows where multiple documents or data sources must be compared. Businesses and professionals using Perplexity’s platform gain a new capability to handle complex, multi-part assignments more efficiently and with improved accuracy.


More in Agents