Perplexity has expanded its Model Council feature to its cloud-based Computer platform, letting users configure multi-model queries with far more choice and customization than the feature offered at launch five months ago.
Model Council, which debuted in February, was originally restricted to three models with no user choice. Now, users can select between two and eight models for their council, drawing from options including OpenAI, Anthropic, Google, and open-weight models like GLM and Kimi. Analysis depth can also be customized, and a synthesizer model puts together the final report.
The expansion turns Model Council from a fixed tool into a flexible skill within Computer, Perplexity's web interface for a model orchestration layer. Perplexity claims the feature can "allow users to build a board of top models for a single issue."
How the council works
Model Council functions as a skill that tells an orchestrator model to spin up parallel subagents. Each model works on the query independently. Perplexity's comms chief, Jesse Dwyer, described the process in detail.
"A wait barrier holds until every perspective finishes, then the model the user selected acts as the ‘chair,’" Dwyer said. The chair model then takes over.
"It reads all the perspective outputs and synthesizes them, explicitly surfacing where the models agree and where they diverge," he added.
Computer can then "turn that synthesis into outputs like reports, board decks, and other work-ready assets," according to Perplexity. The company positions Model Council as ideal for ambiguous questions involving "judgment calls, tradeoffs, assessing risks or other situations where there’s no one right answer."
Intended use cases include legal questions, financial queries, corporate decision making, business growth modeling, and engineering. Users can ask for additional details to "sharpen the answer" and pose pointed follow-up questions.
Consensus and conflict
Perplexity argues the value lies in seeing where models align and where they clash. The company claims model-to-model dialogue "shows where there’s consensus and where there’s conflict."
"Consensus gives people more confidence to move forward," Perplexity said. Gaps, the company argues, show users where to dig deeper.
"Disagreements often come from different starting assumptions, so ask each model to state the assumptions behind its conclusions," Perplexity advised. Users are also encouraged to ask for confidence levels from each model.
Pricing and access
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Access to Model Council was previously restricted to enterprise tiers and individual Max customers. The Computer expansion brings it to individual Pro users, who pay $20 per month. Individual Max costs $200 per month. Enterprise Pro runs $34 per month per seat, while Enterprise Max costs $271 per month.
Computer relies on usage-based billing. Pro users get an initial pool of 4,000 bonus credits, while Max users start with 35,000 and receive 10,000 renewing monthly. Enterprise Pro users get 8,000 initial credits plus 500 per seat each month. Enterprise Max users receive 45,000 initial credits and 15,000 per seat monthly.
Perplexity states that 100 Computer credits equal $1. A light task, like summarizing a document and pulling out key points, can eat between 100 and 350 credits. A heavy task, such as creating pitch decks, can consume as many as 2,275 credits.
Cost uncertainty
The cost of a Model Council session remains unclear. Computer credits function at the product level, not mapping directly to raw AI token usage. Perplexity did not specify how many credits a Model Council session could burn.
Dwyer said query complexity matters more than model choice. "Query complexity will affect token usage more than the specific model selection," he said. "Simple queries in Computer can cost less than a penny while complex tasks run higher."
AI tokens can get expensive, and running multiple models on a single query multiplies the token burn. The article notes it is not clear how many credits a Model Council session could consume, as Perplexity did not get specific.
For users weighing whether to assemble a council of eight models for a legal question or a business growth model, the math is straightforward but the outcome is not. A light task costs between one and three dollars and a half. A heavy task could run past $22. A Model Council session sits somewhere in between, depending on how many perspectives the orchestrator spins up and how long the chair spends synthesizing.
Perplexity's expansion arrives as agentic AI tools proliferate across the industry. The company has been building out its Computer platform with orchestration features, and Model Council now sits inside that stack as a configurable component rather than a standalone offering.
The feature's move to Computer also signals Perplexity's push toward work-ready outputs. Reports, board decks, and other assets generated from multi-model synthesis position the platform as a productivity tool, not just a search engine.
Whether the cost structure holds up for heavy users remains an open question. Perplexity's credit system obscures direct token costs, and the company has not published benchmarks for Model Council sessions specifically.
For now, users can assemble their own councils, choose their chair, and watch the models argue. The price of that argument depends on how complex the question gets.