Sonar Large 32K

Sonar Large 32K is Perplexity’s largest and most capable Sonar model, designed for heavy-duty analysis and retrieval workflows. It extends the R1 reasoning foundation with stronger instruction following, multi-step reasoning, and robust code generation while retaining the lightweight latency profile of the Sonar line. With a 32K context window, it can handle long documents, multiple references, or large codebases without losing coherence, making it well suited for legal and technical document summarization, extended Q&A, and analytics assistants.

Like the rest of the Sonar family, Large 32K is optimized for retrieval-augmented generation. It grounds responses in retrieved evidence, outputs structured formats like JSON for agent use, and can expose its reasoning steps when configured to. Perplexity positions it as the model to choose when you need the most accuracy, context capacity, and reliability in their ecosystem—whether in their Sonar API, the Perplexity app, or third-party integrations. Typical applications include enterprise search copilots, in-depth research agents, repo-level code helpers, and multilingual content analysis at production scale.