Perplexity launches Multiagent effort level for Computer
·2026.11.17 09:00
Key point
The new mode uses a team of agents on different frontier models to work in parallel and cross-check findings for complex tasks.
Details
- Perplexity introduces Multiagent, a new effort level above Ultra for its Computer agent platform, designed for complex work requiring parallel analysis and cross-checking.
- The system uses a lead agent (Claude Fable 5.2) to coordinate up to 16 agents running on different frontier models, including Claude Opus 5.5, GPT 6.1 Sol, and Gemini 4 Carbon.
- Agents share findings directly, check each other's work, and coordinate edits to shared files, allowing the team to identify conflicting assumptions before the lead agent synthesizes a single response.
- Multiagent is available to all Perplexity Max subscribers on the web and is distinct from Ultra (delegated reporting) and Model Council (independent perspectives on the same question).
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.