Perplexity Puts Multiagent Behind Its $167-a-Month Max Plan

Perplexity said on X on Tuesday, 17 November, that it is “rolling out Multiagent, a new effort level in Computer and teammates”. Chief executive Aravind Srinivas posted seven minutes later that it is “Available to all Max users on web”, the plan Perplexity’s pricing page lists at $167 a month when billed annually.

Our number: Max at $167 a month costs 9.8 times the $17 Pro plan, both billed annually, by our arithmetic. Multiagent arrives 105 days after Perplexity added Model Council to Computer and 57 days after it launched Effort Mode, by our arithmetic from its changelog dates.

Business Pill · AI Agents

An evergreen Business Engineer explainer on AI agents, from answering to doing. It teaches the general idea only and does not discuss Perplexity or Multiagent.

The key insight: Perplexity has turned multi-model work from a side-by-side comparison into a division of labour: in Multiagent a lead agent splits one task across agents on different frontier models, which share findings as they go, and Srinivas wrote that it is available to Max users on the web, the $167-a-month plan that costs 9.8 times Pro, by our arithmetic.

What Multiagent Does

Perplexity’s post says “A lead agent splits tasks across a team of agents using different frontier models”. It adds: “As they work, agents share findings with one another, with the lead agent combining everything into one answer”.

Srinivas gave an example: “You can have Fable 5.2 orchestrate Gemini 4 Carbon, GPT 6.1 Sol, Opus 5.5 in parallel agent streams to get complex tasks done faster and better (more accurate)”. That example names four models, by our count.

Srinivas called it “A killer for financial research, forecasting, prediction markets, and analyzing a batch of filings together”. The feature is switched on through the effort setting; Srinivas wrote “toggle on Effort level”.

Perplexity’s plan prices as listed on its pricing page, in dollars a month billed annually: Max $167, Pr
Perplexity’s plan prices as listed on its pricing page, in dollars a month billed annually: Max $167, Pro $17, Free $0. Multiagent is available to Max users on the web, per Perplexity and its CEO on X. The 9.8 times ratio is our arithmetic.

What Was Posted on X

Perplexity’s own announcement, posted at 16:28 UTC on 17 November. It describes Multiagent as a new effort level in Computer and teammates, with a lead agent that splits tasks across agents on different frontier models. The piece quotes it for what the feature does.

Chief executive Aravind Srinivas’s post seven minutes later. It gives the example models, the use cases and the availability: all Max users on web, with mobile support to follow. The piece quotes it for who can use the feature.

How It Differs From Model Council

Perplexity’s 4 August changelog describes Model Council, which lets users “Pick two to eight models across OpenAI, Anthropic, Gemini, and open source”. In that mode “Computer runs each independently and synthesizes where they agree, disagree, and what each uniquely surfaces”.

Multiagent, as Perplexity describes it, works the other way round: the agents “share findings with one another” while they work, and a lead agent splits the task first. Model Council runs the models side by side on one question; Multiagent divides one task among them.

Effort Mode, from Perplexity’s 21 September changelog, let users “Choose Light, Standard, High, or Ultra to match Computer’s effort to the task”. Perplexity now calls Multiagent a new effort level on that same control.

Who Gets It and What It Costs

The pricing page lists Max at $167 a month billed annually, with “10,000 monthly credits” and “35,000 bonus credits”. Pro is $17 a month billed annually, with “4,000 bonus credits”. Srinivas wrote “Mobile support coming soon”.

The business lever is subscribers: Multiagent is a reason to move up to Max, the same tier that got Teammates first, as our Teammates piece reported. Perplexity has also launched Perplexity Monitor, and Srinivas has said the company passed a $1 billion run rate and would not rule out an IPO filing in the first quarter of 2027.

The Structural Read

As we read the two modes, Model Council asks several models the same question and compares the answers, while Multiagent, as Perplexity describes it, hands each agent a part of the task, so the lead agent, not the user, decides who does what.

As we read the packaging, putting Multiagent on the effort control, next to Light, Standard, High and Ultra, makes it a setting a Max user switches on per task rather than a separate product.

Perplexity on X, 17 November 2026

“A lead agent splits tasks across a team of agents using different frontier models”

Three Implications

FOR PERPLEXITY SUBSCRIBERS Multiagent is available to all Max users on the web, Srinivas wrote; Pro, at $17 a month billed annually, is not named in the posts we read.

FOR RESEARCH-HEAVY WORK Srinivas named financial research, forecasting, prediction markets and analyzing a batch of filings together as uses; Perplexity has not published a benchmark for the speed or accuracy claim in the sources we read.

FOR THE MODEL MAKERS Srinivas’s example runs Gemini 4 Carbon, GPT 6.1 Sol and Opus 5.5 under Fable 5.2 as orchestrator, four models reached through one Perplexity subscription in that example, by our count.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Agentic Harness War.

The analysis puts it plainly: “Foundation models become relatively commoditized as orchestration layers capture differentiation.”

As we read it, Multiagent is Perplexity building that orchestration layer in the open: the lead agent picks the work split across rival frontier models, and the subscription sits with Perplexity. The analysis does not discuss Multiagent.

What Is Not Established

Perplexity has not said, in the sources we read, how many credits a Multiagent task uses, how many agents a task can run, or whether users pick the models. Srinivas’s four-model line is an example, not a stated limit.

The claim that tasks finish “faster and better (more accurate)” is Srinivas’s; we found no benchmark for it in the sources we read. Perplexity’s changelog had no Multiagent entry when we read it at about 16:56 UTC. We did not contact Perplexity.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

The Agentic Harness War

The Business Engineer analysis of how the agentic surface, the harness, is becoming where value pools in the AI stack.

Read the Map of AI →

The Bottom Line

Perplexity’s Multiagent lets a lead agent split one task across agents running different frontier models, and it is available to Max users on the web. Max costs $167 a month billed annually, 9.8 times Pro, by our arithmetic.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. Multiagent’s description is from Perplexity’s post on X and the availability, example models and use cases are from Aravind Srinivas’s post on X, both on 17 November 2026; Perplexity had not published a changelog entry on Multiagent when we read its changelog. Plan prices and credits are from Perplexity’s pricing page; Model Council and Effort Mode are from its changelog. The 9.8 times, 105-day and 57-day figures are our arithmetic. We did not contact Perplexity.

Sources: Perplexity on X, 17 Nov 2026 · Aravind Srinivas on X, 17 Nov 2026 · Perplexity pricing page · Perplexity changelog, 21 Sep 2026 · Perplexity changelog, 4 Aug 2026

Scroll to Top
FourWeekMBA