Perplexity

Perplexity sells four developer APIs for web search and web grounded answers: a Router API, an Agent API, a Search API and an Embeddings API. Requests authenticate with a bearer token supplied in the Authorization header, and billing is pay as you go, with no subscription required.

The Agent API returns answers grounded in the web with citations attached, and it can also route a request to third party models from OpenAI, Anthropic, Google and xAI. The Search API is the lower level option, returning ranked raw web results rather than a written answer. Each result carries a title, a URL and a content snippet, with publication and last updated dates available as optional fields. A call accepts between one and twenty results and returns ten by default, can be restricted to at most twenty domains and twenty languages, and can be filtered by publication date, by last updated date, or by a recency window of an hour, day, week, month or year. The Router API puts open weight models hosted by Perplexity behind a single OpenAI compatible endpoint.

Pricing differs by product. The Search API is 5.00 USD per 1,000 requests. Sonar is billed at 1 USD per million input tokens and 1 USD per million output tokens plus a per request fee, and Sonar Pro at 3 USD per million input tokens and 15 USD per million output tokens plus a per request fee. Agent API tool calls are billed per invocation: 0.0025 USD for a web search, 0.0005 USD for fetching a URL, and 0.005 USD each for people search and finance search. Third party models routed through the Agent API are billed per million tokens. Embeddings are cheaper again. The pplx-embed-v1-0.6b model returns 1024 dimensions, accepts up to 32K tokens and costs 0.004 USD per million tokens, while pplx-embed-v1-4b returns 2560 dimensions at 0.03 USD per million tokens.