Perplexity Brings Multi-Model 'Council' Feature to Computer Platform, Lets Users Pick Up to 8 AI Models

Perplexity on Tuesday expanded its Model Council feature to its Computer platform, letting users select between two and eight AI models to independently analyze a single query before a synthesizer model reconciles their conclusions into one report.

Model Council debuted in February as a fixed three-model synthesis tool with no user control over which models participated. The Computer integration changes that: users can now build a custom "board" from models supplied by OpenAI, Anthropic, Google, and open-weight options including GLM and Kimi.

How it works

According to Perplexity communications chief Jesse Dwyer, Model Council operates as a skill that instructs an orchestrator model to spin up parallel subagents, each functioning as a full agent harness working the problem independently. A "wait barrier" holds until every model finishes its analysis. Then a user-selected "chair" model reads all outputs and synthesizes them, explicitly flagging where the models agree and where they diverge.

Perplexity positions the tool for "ambiguous" questions involving judgment calls, tradeoffs, or risk assessment—cases with no single correct answer. Suggested use cases include legal analysis, financial queries, corporate decision-making, business growth modeling, and engineering problems. The company recommends asking the Council to state confidence levels and underlying assumptions, noting that "disagreements often come from different starting assumptions."

With the Computer integration, users can also ask follow-up questions to refine results and convert the synthesis into formatted outputs like reports and board decks.

Pricing and credits

Model Council was previously gated to Perplexity's enterprise tiers ($34/month per seat for Pro, $271 for Max) and individual Max subscribers ($200/month). It's now also available to individual Max and Pro ($20/month) users.

Computer runs on usage-based credit billing rather than direct token pricing. Perplexity says 100 credits equal $1. Pro subscribers get 4,000 bonus credits initially; Max users get 35,000 upfront plus 10,000 renewing monthly. Enterprise Pro seats get 8,000 bonus credits plus 500 monthly; Enterprise Max seats start with 45,000 and get 15,000 per seat monthly.

Perplexity says a "light" task—like summarizing a document—costs 100 to 350 credits, while a "heavy" task, such as generating a pitch deck, can consume up to 2,275 credits. The company did not disclose how many credits a full Model Council session with multiple models might burn, since credits map to product-level tasks rather than raw token counts. Dwyer said "query complexity will affect token usage more than the specific model selection," adding that simple queries can cost less than a penny while complex tasks run higher.

What this means

Model Council is Perplexity's bet that showing disagreement between AI models is more useful than hiding it behind a single answer—a reasonable pitch for high-stakes, ambiguous business questions where a false sense of certainty is the real risk. But running up to eight full agent harnesses in parallel is expensive by design, and Perplexity's refusal to specify credit costs for a Council session makes it hard for users to budget for the feature before they use it. Given that Computer already charges per credit consumed beyond bonus pools, enterprises drawn in by the promise of multi-model consensus should expect metered costs that scale directly with how many models they add to the board and how deep they push the analysis.