Model Council comes to Computer
Model Council comes to Computer
Every new release in AI pushes frontier models further into their own specialties and strengthens the case for a multi-model approach.
In practice, that means either using the best model for each task or weighing several models’ perspectives on a single issue. Over the course of a single Computer task, those judgement calls can surface repeatedly.
Today we’re rolling out Model Council in Computer. Model Council allows users to build a board of top models for a single issue.
When a user selects Model Council, they now pick the lineup themselves: two to eight models from OpenAI, Gemini, Anthropic, and open source providers. They can also select the analysis depth, ranging from brief insights to detailed reports. Model Council synthesizes where the models agree, where they disagree, and what each one uniquely surfaces. Now it can also use Computer’s capabilities to turn that synthesis into outputs like reports, board decks, and other work-ready assets.
The harness improves as models get better
One of top reasons people choose Perplexity is they’re never locked into one model or model family. Perplexity is model agnostic and always adding the latest frontier models to its harness.
Perplexity gets better as the models get more powerful, because it assigns the right model to each task whether it’s research, analysis, coding, or asset generation.
The same idea drives Model Council. Pitting the top models against each other shows where there’s consensus and where there’s conflict. Consensus gives people more confidence to move forward. Gaps show them where to go deeper.
Get rich perspective on high-stakes decisions
People bring Model Council all sorts of challenging questions. Here are a few examples:
Legal: “Review this indemnification clause and rank the top risks to the party accepting it. What are we missing? Create a PDF I can share with my team.”
Finance: “Act as a partner at a VC. How should I value a pre-revenue clean energy hardware startup raising a $15M Series A at a proposed $75M post? Give me the confidence level with your conclusion, then create a deck I can share with the board.”
Corporate strategy: “Act as a CFO. We need a modern warehouse management system within 12 months. Should we build or buy given that timeline? Explain the assumptions behind your conclusion, then prepare a memo and strawman Excel sheet with sample options.”
Personal: “What are the tradeoffs between a high-deductible HSA plan and a low-deductible PPO for a family of four with high medical usage? Create a Google doc I can share with my spouse.”
Popular Model Council use cases
Industry | What people ask for |
|---|---|
Legal | Case arguments, statutory questions, landlord-tenant issues, immigration decisions, and inheritance options |
Finance | Earnings comparisons, valuations, tax treatments, and portfolio construction |
Corporate | Reasoning over files, drafting structured emails, and long-form article fact checking |
Growing Business | Pricing models, market entry, competitive positioning, and operating trade-offs |
Engineering | Code drafting, architecture evaluation, and library comparisons |
Access Model Council in different ways
There are two ways to start a Model Council query. Select Model Council from the dropdown menu when selecting Computer in the omnibar or type “run a Model Council” directly in the query.
From there, Model Council offers a choice of which models to include in the answer, among them Claude Opus 4.8, ChatGPT 5.5 and Gemini 3.1 Pro (as of late July 2026).
Model Council is built for ambiguous questions: judgment calls, tradeoffs, assessing risks or other situations where there’s no one right answer.
Asking for decisions, not summaries
Model Council is even more powerful in Computer because it goes a step beyond research. It runs the same multi-model debate, then uses Computer to turn the synthesis into an accurate, work-ready output.
The more pointed the question, the stronger the result. Ask Model Council what it would bet on and why. Pit models against each other with questions like “where would a smart person disagree with you?”
Details sharpen the answer, so include numbers, timelines, budget, and tradeoffs. Assigning a persona helps too, such as “answer as a manager” or “as a CFO.” To get perspectives on a document or contract, upload it and query Model Council directly.
For the synthesized answer, ask for confidence levels. When one model lands at 60% and another at 95%, that gap reveals uncertainty. Disagreements often come from different starting assumptions so ask each model to state the assumptions behind its conclusions.
Now available to all Computer users
We build for curious people. Nearly half of all Model Council users are executives, the people most often saddled with hard decisions.
But the feature is for anyone who wants to know not only when they’re right, but also what they’ve overlooked.
Model Council is available on the web and mobile to all Max users and Pro users in Computer.
Try Model Council in Computer today.