Perplexity adds a Multiagent level to Computer, where AI agents on different models check each other
It is available to Perplexity Max subscribers on the web. The capabilities described are Perplexity's own.
On 17 November 2026 Perplexity launched Multiagent, a new effort level for its Computer agent platform, above the existing Ultra setting. Perplexity says it is meant for complex work with several connected parts.
How it works, per Perplexity
A lead agent plans the task and runs a team of agents on different frontier models. They work in parallel, share findings directly, and check each other's work. Perplexity says this helps surface conflicting assumptions before everything is combined into one answer. The lead can assign up to 16 agents. The default team combines Claude Opus 5.5, GPT 6.1 Sol and Gemini 4 Carbon, and users can name the models they want.
Examples and availability
Perplexity's examples include financial forecasting, annual report review, deal data room diligence and year-end tax planning, where it says the output is meant to support questions for an auditor or tax professional. Multiagent is available to all Perplexity Max subscribers on the web, selected from Computer's effort slider.
What to keep in mind
This is a vendor announcement. Perplexity gives no benchmark results for Multiagent in the post, and we have not tested it. Cross-checking between models can catch some errors but does not guarantee correct answers, so important numbers still need human review.