Perplexity Unveils "Model Council": A Supreme Court for AI Answers

On February 7, 2026, Perplexity launched Model Council, a breakthrough feature that eliminates "model hopping" by running a single query across the world’s leading AI systems simultaneously.

​How the Council Works:

​Parallel Processing: One prompt is instantly distributed to Claude Opus 4.5, GPT-5.2, and Gemini 3.0. The system draws on the unique reasoning strengths of each model.

​The "Synthesizer" Layer: A specialized meta-model analyzes the raw outputs to create a unified briefing. It highlights "Consensus Points" where models agree and flags "Outliers" where they differ.

Fact-Checking through Redundancy: By cross-referencing multiple frontier models, Model Council significantly lowers the risk of hallucinations, making it an essential tool for high-stakes research and legal/medical inquiries.

​User-Driven Benchmarking: Users can see which model provided the most comprehensive answer, effectively creating a real-time leaderboard for their specific use cases.

​Availability: Now live for Perplexity Max subscribers on the web, with mobile integration coming in the next update.

​Perplexity is shifting from being an "answer engine" to a "reasoning orchestrator," providing users with the safety and depth of a collective AI intelligence.