Model Selection and Model Council

Perplexity Is Not One Model

When you search on Perplexity, an AI model synthesises the answer — but Perplexity does not have just one model. It runs several frontier models from different providers, and you can choose which one handles your query. This is unusual in consumer AI products and gives Perplexity a meaningful advantage: when a new frontier model launches, Perplexity can add it within days.

Available Models

The model lineup evolves as providers release updates, but as of 2026 the options inside Perplexity include:

• Sonar — Perplexity's own model, optimised for web-grounded search. Fast, efficient, and the default for most queries. Best for everyday factual lookups.
• Claude (Anthropic) — Strong analytical reasoning, nuanced writing, and careful handling of complex documents. Best for writing tasks, document analysis, and anything requiring precise, detailed prose.
• GPT (OpenAI) — Versatile and strong at structured tasks, coding explanations, and step-by-step reasoning.
• Gemini (Google) — Deep integration with Google's knowledge graph; strong at multilingual queries and content connected to Google's data ecosystem.
• Grok (xAI) — Good at recent events, direct answers, and culturally current content.

The default (Auto) mode lets Perplexity choose the model for each query. For most users, this is the right setting — Perplexity matches the model to the task type.

When to Override the Default

Switch models manually when you have a specific reason:

• Writing a polished document → Claude for prose quality
• Explaining a coding concept step by step → GPT
• Asking about a very recent news event → Grok
• A query in a language other than English → Gemini

Model Council: Compare Answers Side by Side

Model Council is a Perplexity feature that runs your query through multiple models simultaneously and shows you each answer in a side-by-side view. You can compare how Claude, GPT, and Grok each respond to the same question — useful when you want multiple perspectives, when the topic is contested, or when you are not sure which model will give the best result.

To access Model Council, look for the option in the model selector before submitting your query. It uses more credits than a single-model search but can save the time of running the same query multiple times manually.

A Practical Note on Credits

Frontier models (Claude, GPT, Grok, Gemini) cost credits on Pro plans. The default Sonar model does not. If you are on a free or Pro plan and want to preserve your credit balance, use Auto mode for most queries and switch to a specific frontier model only when the task genuinely benefits from it.