Perplexity Launches Model Council to Compare Multiple AI Models in Parallel.

Perplexity has launched Model Council, a new feature that runs queries through three frontier AI models simultaneously and synthesizes their outputs into a single, more reliable answer. The capability is available now on the web for Perplexity Max subscribers.

Model Council addresses a core challenge in AI-powered search: different models excel at different tasks, and no single model consistently delivers the best answer. By running queries in parallel across three selected models, users can compare how each interprets the question and what evidence each surfaces.

The system works in three steps. First, it sends the user's query to three chosen AI models simultaneously. Second, it displays the responses side by side so users can review each model's reasoning and sources. Third, a separate synthesis model reviews all three outputs, highlighting where they agree, where they diverge, and what unique insights each contributes. The final answer reflects this consensus view, flagging areas of uncertainty or disagreement.

Model Council targets users who need confidence in their answers, particularly for research, analysis, or decision-making where accuracy matters. Rather than accepting a single model's perspective, users can see which claims are corroborated across models and which rest on shakier ground.

The feature builds on Perplexity's existing multi-model infrastructure. As previously reported, the company has invested heavily in optimizing model serving at scale. Model Council extends that capability, allowing subscribers to leverage multiple frontier models without switching interfaces or re-running queries manually.

Perplexity did not specify which models are available for selection, but the term "frontier models" typically refers to the most advanced offerings from labs like OpenAI, Anthropic, and Google. Users can presumably choose any three models from the available pool, tailoring the Council to their specific query type.

Model Council is limited to Perplexity Max subscribers, the company's premium tier. This follows a pattern of gating advanced features behind the paywall; recent updates have similarly reserved premium data sources for paying users.

The web-only restriction suggests the feature is computationally intensive. Running three models in parallel, then synthesizing results with a fourth, likely increases latency and cost compared to single-model queries. Perplexity may expand availability to mobile apps or other tiers depending on adoption and infrastructure capacity.

The move positions Perplexity as a meta-layer atop the AI landscape, rather than a single-model product. While competitors like ChatGPT and Claude offer direct access to their respective models, Perplexity is betting that orchestrating multiple models, comparing their outputs, and surfacing consensus answers will deliver a better user experience for complex queries.

Model Council reflects a broader shift in how AI products are positioned. Instead of claiming one model is definitively better, the approach acknowledges trade-offs and gives users transparency into how answers are constructed. For business owners and researchers using AI for high-stakes work, seeing where models agree and disagree can be more valuable than a single confident answer that might be wrong.

The feature also raises questions about cost and speed. Parallel inference is expensive, and synthesis adds another layer of processing. Perplexity has not disclosed pricing changes or query limits for Model Council, but Max subscribers should expect slower response times compared to standard queries.

Model Council is live now at perplexity.ai for Max subscribers on the web.