Perplexity Model Council Expands: Pick 2-8 AI Models for One Answer

Perplexity on Tuesday rolled out an expanded version of its Model Council feature to its Computer platform, five months after the tool first launched. Where the original version quietly picked three models with no user input, the new setup lets users choose 2 to 8 models from a mix of frontier providers and open-weight options, run them in parallel on a single question, and get a synthesis report that flags where the models agree and where they split. Access is now open to $20/month Pro subscribers for the first time, though usage-based billing means complex tasks can add up fast.

What happened

Detail	Fact
Launch date	Tuesday, July 29, 2026
Original Model Council launch	February 2026
Model range	2 to 8 models per session
Model sources	OpenAI, Anthropic, Google, GLM, Kimi (open-weight)
Pro plan price	$20/month (individual)
Max plan price	$200/month (individual)
Credit rate	100 credits = $1
Light task cost	100 to 350 credits
Heavy task cost	Up to 2,275 credits

Perplexity’s Model Council works as a skill layered on top of an orchestrator model. According to comms chief Jesse Dwyer, the orchestrator spins up parallel subagents, each acting as a full agent that works the problem independently. A wait barrier holds everything until every model finishes, then the user’s chosen “chair” model reads all the outputs and writes a synthesis that explicitly calls out where models agree and where they differ.

Computer, Perplexity’s cloud-based platform, is essentially a web interface for that orchestration layer. Adding Model Council to it means the synthesized output can go further: rather than just a summary, Computer can turn the result into reports, slide decks, and other finished documents.

How the pricing actually works

Credits are tied to the Computer product, not to raw token counts, so there is no direct way to calculate cost per model call. Dwyer told The Register that query complexity affects token usage more than which specific models are selected, and that simple queries can cost less than a penny while complex tasks run higher.

Plan	Bonus credits (one-time)	Recurring credits/month
Pro (individual)	4,000	Not specified
Max (individual)	35,000	10,000
Enterprise Pro (per seat)	8,000	500
Enterprise Max (per seat)	45,000	15,000

The previous version of Model Council was locked behind enterprise tiers ($34/month per seat for Enterprise Pro, $271 for Enterprise Max) and the individual Max plan. The Computer expansion is the first time individual Pro subscribers can use it.

Why it matters

Running several models against the same question and comparing their conclusions is a practical way to pressure-test a decision. Perplexity specifically positions the tool for situations with no single correct answer: legal questions, financial analysis, corporate decisions, business growth modeling, and engineering trade-offs. When models reach the same conclusion by different paths, that convergence is a meaningful signal. When they split, you know where more research is needed.

The synthesis-to-document capability in Computer is worth noting for business users. Getting a structured output like a board deck or a risk report directly from a multi-model session compresses what would otherwise be several manual steps. For teams already using AI integration in their workflows, this kind of orchestrated output reduces the gap between “AI gave me an answer” and “AI gave me something I can act on.”

The broader trend here also matters: we have covered how model routing and cost control are becoming central concerns as AI usage scales. Perplexity’s credit system abstracts away token costs, which makes budgeting easier but also less transparent.

Our take

The Model Council concept is genuinely useful for ambiguous, high-stakes questions. Running GPT-4o, Claude, and Gemini on the same legal or financial question and reading where they diverge is better due diligence than trusting one model’s confidence. The “chair” synthesis framing is smart: it forces an explicit summary of disagreements rather than blending them into a falsely unified answer.

That said, the credit pricing is opaque. A heavy task at up to 2,275 credits costs $22.75 per session at the stated rate. Run a few complex Model Council queries a week and a Pro account’s bonus credits are gone quickly. Perplexity did not specify how many credits a typical multi-model Council session consumes, which makes it hard to plan. Watch your credit balance closely before running this on anything lengthy.

Perplexity’s decision to push this down to the $20/month tier is smart positioning against ChatGPT and Claude’s comparable plan prices, but the value is only there if the credit economics work out. Test it on a smaller question first before routing your actual board-level decisions through it.

What to do about it

1. Check whether your current Perplexity plan (Pro, Max, or Enterprise) has been updated to include Computer access with Model Council.
2. Start with a “light” task (100 to 350 credits) to calibrate what a session actually costs before running heavy document-generation tasks.
3. For important decisions, ask each model to state its assumptions explicitly, as Perplexity recommends. Disagreements rooted in different assumptions are more useful than ones that look like random noise.
4. If you need polished outputs like pitch decks or risk reports from multi-model analysis, test Computer’s document generation before committing it to a client-facing workflow.

If you want help deciding where multi-model AI fits into your existing business processes, the Lumien team is happy to talk it through.

Source: The Register · AI/ML