Perplexity Expands Decision-Support Tool "Model Council" to Utilize Up to 8 Models

AI search company Perplexity has expanded its “Model Council” feature, which integrates the results of multiple large language models (LLMs) running in parallel, to its cloud execution environment “Computer.” This allows users to form an “AI advisory board” with any combination of 2-8 models, ranging from cutting-edge closed models like those of OpenAI, Anthropic, and Google, to open-weight models such as GLM and Kimi.

Background:

From Limited Release to Full Deployment

Model Council was first introduced in February 2026, initially limited to three models available on Perplexity, with no option for users to select the models. When users submitted queries, the synthetic model collected responses from idle models and generated a consolidated report. With its expansion to the Computer platform, users now have significantly greater freedom in choosing models.

According to The Register’s Brandon Vigliarolo, Model Council functions as a “skill,” instructing an orchestrator model to launch parallel sub-agents. Jesse Dwyer, Perplexity’s head of communications, explained, “Each sub-agent independently tackles the problem at hand, held by a wait barrier until all perspectives are complete. Then, the user-selected model serves as the ‘chairperson,’ reading each output and producing a synthetic report that explicitly highlights points of agreement and disagreement.”

Integration with the Computer Platform

Computer, Perplexity’s cloud execution environment launched in 2025, serves as a web interface for model orchestration. Users can connect multiple AI models via their browser to execute tasks. The addition of Model Council further enhances Computer’s functionality.

Specifically, when using Model Council on Computer, users can customize the depth of analysis. In addition to generating reports, the synthetic model can transform synthesis results into business assets such as reports and executive materials. Perplexity also stated that users can refine results by requesting additional details or posing focused follow-up questions, thereby achieving more robust outcomes.

Mechanism of Multi-Model Synthesis

According to Dwyer, Model Council first activates an orchestrator model to launch multiple sub-agents in parallel as “skills.” Each sub-agent operates as a fully independent agent harness, performing reasoning tasks on the given problem. A wait barrier ensures all sub-agents complete their tasks, after which the user-selected model acts as the chairperson to read all outputs.

The chairperson model’s role goes beyond mere summarization. It compares each model’s output, explicitly identifying points of consensus and divergence. Through this process, users can “proceed confidently where consensus exists and identify areas for deeper exploration where gaps remain,” Perplexity explained.

Anticipated Use Cases

Perplexity highlights several potential applications for Model Council, including legal consultations, financial queries, corporate decision-making, business growth modeling, and engineering. The tool is particularly suited for ambiguous problems without a single correct answer—situations requiring judgment on trade-offs, risk assessments, and other scenarios. By visualizing consensus and points of contention among models, Model Council offers valuable insights to human decision-makers.

Differentiation from Competitors

In the AI search market, Perplexity has long sought differentiation through simultaneous utilization of multiple models. With the expanded Model Council, the company is moving beyond a mere search engine to position itself as an enterprise-grade decision-support platform. By enabling users to access diverse perspectives from various AI models, Perplexity sets itself apart from competitors reliant on single models.

Additionally, since Model Council operates on the Computer platform, users can access computational resources as needed via the cloud. The Register has praised this development, noting that “running up to eight models in the cloud has become affordable.”

Editorial Opinion

The expansion of Model Council represents a clear strategic move by Perplexity to penetrate the enterprise market. As the number of AI models continues to grow, the company has addressed the challenge of determining which model to trust by proposing a “consensus-based” solution. This approach is commendable and could lower barriers to AI adoption in high-accountability fields such as risk assessment and legal decision-making, where transparency is critical.

In the long term, the process of visualizing agreement and disagreement among models could democratize the concept of “AI ensemble.” While currently limited to cutting-edge models, the inclusion of specialized and domain-specific models in the future could pave the way for genuine multi-model decision-making. However, it is worth noting that consensus among models does not always equate to correctness; there is a risk of collective bias emerging.

One question for the editorial team is how users will practically utilize synthesis results from up to eight models, particularly considering cognitive load. Further evaluation from this perspective is warranted.

References

• “Perplexity’s tokenmaxxing Model Council gives you multiple bot perspectives”, by Brandon Vigliarolo — The Register, 2026-07-28T21:26:53.000Z (ARR)
• Source URL: https://www.theregister.com/ai-and-ml/2026/07/28/perplexitys-tokenmaxxing-model-council-gives-you-multiple-bot-perspectives/5279972