Perplexity Launches Multiagent Mode With 16 AI Agents
A new effort level for Computer
Perplexity launched a Multiagent effort level for its Computer product on November 17, 2026, positioning it above the existing Ultra tier and making it available on the web to Perplexity Max subscribers (Perplexity). The system pairs a lead AI agent with as many as 16 additional agents drawn from Anthropic's Claude, OpenAI's GPT and Google's Gemini model families (dailyaibrief).
Why it matters for agent builders
For people building or buying agent systems, Multiagent is a concrete test of whether orchestrating multiple models from different vendors inside a single task can be packaged as a consumer-facing product rather than a custom engineering project. Perplexity's design — one lead agent splitting a complex task into subtasks, assigning them across up to 16 workers, then having those workers cross-check each other before a final answer is produced — mirrors patterns enterprise teams have been assembling manually with frameworks and APIs. Shipping it as a one-click effort level inside an existing subscription product signals that multi-model, multi-agent coordination is moving from bespoke pipelines toward productized tooling.
How the delegation works
According to Perplexity's own description, the lead agent "coordinates AI models to analyze complex tasks and cross-check their work" (Perplexity). In practice, the lead agent can hand off pieces of a research task to the pool of up to 16 agents while also working on part of the task itself. Those agents operate on separate components of the problem in parallel, share what they find with one another, and review each other's output before the lead agent assembles a single response for the user (Emergent).
What's confirmed and what isn't
Perplexity has not published performance data, accuracy benchmarks, or latency figures comparing Multiagent to its Ultra effort level or to single-model queries, and no model version numbers for the Claude, GPT or Gemini systems involved have been disclosed (i10x). The launch is confirmed as a web feature for Max subscribers; reporting on broader mobile availability has circulated but the primary announcement centers on the web rollout (Daily AI Brief).
What comes next
The launch puts Perplexity among a small group of consumer AI products explicitly marketing cross-model agent orchestration rather than a single proprietary model stack. Whether the 16-agent ceiling, the cross-checking step, or the specific model mix changes will depend on usage data Perplexity has not yet released.


