Perplexity Sonar Pro

Perplexity Sonar Pro is built as an enterprise search and reasoning API that layers language model reasoning over live information retrieval. The model accepts both text and image inputs to let users ground queries in uploaded documents, and it returns text responses enriched with citations drawn from current sources. Its larger context window enables in-depth, multi-step query handling and supports longer, more nuanced follow-up conversations compared to the base Sonar tier, making it better suited for complex research workflows where breadth and depth matter simultaneously.

Sonar Pro is designed for developers and teams that need authoritative, sourced answers delivered at scale, with a zero-retention data policy that keeps prompts out of training and logging systems. Performance characteristics reported through OpenRouter show throughput around 72 tokens per second and end-to-end latency averaging roughly 1.78 seconds, positioning it as a practical choice for production applications requiring both depth and responsiveness. By doubling citation counts per search on average, the model helps users trace insights back to original sources more efficiently, fitting well into use cases ranging from internal research tooling to customer-facing products that demand verifiable, up-to-date information.