Perplexity: Sonar Deep Research

Perplexity Sonar Deep Research is a text-only model from Perplexity with a 128,000-token context window. It supports reasoning, which positions it for multi-step analytical tasks, but it does not support tool use and has no structured output capability. There is no published maximum completion length in the available specs, which is worth confirming before committing to long-form generation workflows. On the comparison question, the $2.00 per million input and $8.00 per million output pricing sits in a mid-range tier, and the reasoning support may appeal to teams running research synthesis or document analysis tasks. The significant caveat is that no independent benchmark coverage exists for this model yet, so there is no third-party performance data to weigh against that cost. Buyers who need validated accuracy metrics before deployment should treat it as unproven and consider waiting for coverage to emerge.