Sonar Reasoning Pro

Sonar Reasoning Pro sits within Perplexity's Sonar family as the dedicated reasoning tier, distinct from the base Sonar and Sonar Pro offerings and oriented toward multi-step answer generation that benefits from live web grounding. Perplexity introduced it alongside Sonar Pro in a March 2025 improvement announcement that framed the family as search-augmented models designed to combine reasoning with up-to-date retrieval. In that same release, Perplexity reported its own benchmark testing showing Sonar and Sonar Pro outperforming search-enabled GPT-4o, and grouped Sonar Reasoning Pro with Sonar Pro as the premium tiers eligible for simplified, citation-aware billing. The intended use case is therefore complex, search-dependent queries where users want both transparent reasoning and fresh source-backed answers rather than a fast, lightweight chat reply.

Operationally, Sonar Reasoning Pro inherits the same tiered search mode controls Perplexity rolled out for the broader Sonar lineup: High for maximum depth and context on complex queries, Medium as a balanced default, and Low for cost-efficient straightforward questions, all applied per request rather than fixed at the model level. It is also covered by Perplexity's policy of not charging for citation tokens, making it practical for workflows that surface many inline references. Because the reasoning capability is paired with retrieval, it fits research assistants, analytical dashboards, and any product that needs explainable, source-cited answers to questions requiring both inference and current information, while teams that only need lightweight chat can lean on the lower Sonar tiers instead.