Perplexity launches the Search API today, September 25, 2025, opening its web-scale retrieval stack to developers. The service exposes an index spanning hundreds of billions of pages and returns ranked snippets crafted from sub-document chunks. Target users are startups, enterprises, and agent builders. Availability is public via the API Platform with SDKs for Python and TypeScript. Pricing is $5 per 1,000 requests with no token fees for Search API calls.
The API supports regional targeting by ISO country code, date range filters, and domain allowlists or denylists up to 20 entries. Multi-query requests bundle up to five queries. Results per call can be set from 1 to 20, default 10. Content extraction is tunable through a max_tokens_per_page setting that defaults to 1024. Responses include title, URL, snippet, publication date, and last_updated fields. An academic mode prioritizes scholarly sources. Rate limits scale by usage tiers tied to cumulative spend.
Perplexity claims lower cost, strong latency, and higher relevance, and releases an open evaluation framework to validate comparisons. The API Platform also unifies Search with Sonar chat models for grounded answers when generation is needed. Early customer stories feature Zoom, Copy.ai, and Doximity, with developer events scheduled in San Francisco and London next month.
Perplexity positions the launch as part of its accuracy-first approach, citing real-time indexing that processes tens of thousands of updates per second and a policy of not training LLMs on customer data. The company is expanding from a user-facing answer engine to a developer platform aimed at powering agentic and search-driven products at global scale.
