Skip to main content

Overview

sonar is Perplexity’s live web-search model. Instead of answering from training data alone, it runs a real-time search of the web on every request and returns a current answer with citations. Reach for it when the answer depends on fresh information — news, prices, scores, releases, schedules, or docs that change after a model’s training cutoff. Sonar does its own searching internally. You send a question; it returns a grounded, cited answer. There is no tool-calling step for you to wire up — and no tool-calling support if you wanted one (see below).

Specs

Pricing

Sonar bills two ways on the same request:
  • Tokens1.35/1Minput,1.35 / 1M input, 1.35 / 1M output, just like any chat model.
  • Web-search fee — a flat $0.00675 per request, on top of tokens, because every call performs a live search.
Short questions are dominated by the search fee, not the tokens. Every response’s usage.cost already includes the search fee. See Pricing → Web search for the full explanation.

Use this when

  • The answer depends on current information — news, prices, releases, schedules, live status.
  • You want sources: Sonar returns citations alongside the answer.
  • You want web grounding without building your own search + scrape + RAG pipeline.

Pick something else when

  • You need deeper, multi-step research or longer cited reports: use sonar-pro.
  • You need function calling / tools: Sonar has none — use kimi-k2.6 or qwen-3.6-plus.
  • The task does not need fresh web data: a standard model is cheaper, with no per-request search fee.

Example