Source: Perplexity API docs (Search API, Agent API models and effort levels)

Perplexity's developer docs as of late 2026: the raw Search API priced per request, the Sonar grounded model, and Agent API effort presets that bundle a model, step budget and web tools.

Contents

Summary

Perplexity offers three layers: a Search API returning ranked results, a grounded answer model (perplexity/sonar), and an Agent API that runs third-party frontier models with Perplexity’s web tools under named effort presets.

Key claims

  • Search API: POST /search, no model parameter, returns ranked results; $5 per 1,000 requests, or $1 per 1,000 with search_type: "fast"; up to five queries in one request bill as one unit.
  • perplexity/sonar (grounded search model): $0.25 input / $2.50 output per million tokens (2026-11). The current model page lists no other Sonar variants (the older sonar-pro, sonar-reasoning-pro and sonar-deep-research names do not appear).
  • Agent API effort presets: fast (1 step, web_search), low (5 steps, adds fetch_url), medium (15 steps), high (40 steps), xhigh (300 steps, adds sandbox and finance search, Claude Opus 5.5 by default), ultra (broad coverage). Presets can be pinned by version or frozen into a request.
  • The effort preset is separate from the model’s own reasoning.effort.

Feeds

Web search APIs, Deep research agents

Suggest changes →