Perplexity Lets You Build a Panel of AI Models to Debate Your Hardest Questions
The search company's Model Council feature now lets paying users pick up to eight AI models, have them tackle a question independently, then read a synthesis of where they agree and where they split.
Key points - Perplexity expanded its Model Council feature to its Computer platform on Tuesday in June 2025. - Users can now select two to eight AI models to analyse a question in parallel, up from a fixed three with no user choice at launch. - The feature is now available to individual Pro subscribers at $20 a month, previously only accessible from $200 a month upward. - Credits cost $1 per 100, and a single heavy task can consume up to 2,275 credits. - Model Council is designed for questions with no single right answer, such as legal, financial or strategic decisions.
Perplexity, the AI-powered search company, has given its Model Council feature a significant upgrade, and for the first time it's available to subscribers paying as little as $20 a month.
Model Council, first launched in February 2025, works like a panel discussion between AI models. You pose a question, multiple models each work on it separately, and then a "chair" model reads all their answers and writes a single synthesis report showing where the models agree and where they differ. Think of it as asking four consultants the same question, then having a fifth summarise the combined advice.
What exactly changed?
The update brings Model Council into Perplexity's Computer platform, a web-based tool that sits on top of a model-orchestration layer, meaning software that coordinates multiple AI models working together behind the scenes. Previously, the Council ran with three pre-selected models and no user input. Now users can pick two to eight models, choosing from big-name providers including OpenAI and Anthropic, as well as open-weight models (freely available models anyone can download and run) such as GLM and Kimi.
When we covered the Computer platform's Windows launch on 28 July, the tool cost $200 a month and was aimed squarely at business users. Dropping access to $20 a month is a real shift.
Computer can also convert the synthesis into finished documents: reports, board decks or other work-ready files, not just a block of text.
Who pays how much?
| Plan | Monthly price | Bonus credits | Monthly credits |
|---|---|---|---|
| Individual Pro | $20 | 4,000 | Not specified |
| Enterprise Pro | $34/seat | 8,000/seat | 500/seat |
Credits are the currency Computer uses. One hundred credits equal $1. A light task such as summarising a document costs roughly 100 to 350 credits. A heavy task like producing a finished pitch deck can hit 2,275 credits in a single session.
Perplexity didn't confirm how many credits a full Model Council session burns. Comms chief Jesse Dwyer told The Register that query complexity matters more than which models you pick: "Simple queries in Computer can cost less than a penny while complex tasks run higher." Helpful context, but costs are hard to predict before you run a session. Our 18 July story on Google's new Gemini credit rules found the same problem: vague billing surprises users who assumed their allowance would go further.
Should you try it?
Model Council is genuinely well-matched to questions without clean answers: a business owner weighing a new market, a freelancer reviewing contract terms, a manager stress-testing a hiring plan. Seeing where models disagree can be as useful as seeing where they converge, because disagreement usually flags hidden assumptions worth examining.
But watch the credits. Run a few light test queries first before you feed the Council your most complex strategic question.
Here's the honest caveat: users most likely to share success stories are the ones who got clean, useful answers. AI models still hallucinate, meaning they confidently state things that aren't true, and a synthesis of several wrong answers is no better than one. Treat the output as a starting point, not a final verdict.
GLM, one of the selectable open-weight models here, is the same model family our 21 July story flagged as cheap but prone to hallucinations. Worth keeping in mind if you're building a council that includes it.
Takeaway: Start with a $20 Pro subscription, use the 4,000 bonus credits on a real but non-critical decision, and see whether the multi-model debate adds anything your own judgement would have missed.
