Exa vs Perplexity Search API

The boundary

For standard search requests with no extra results, page contents, or summaries, Exa is cheaper below 5,000 requests a month, the bills tie at 5,000, and Perplexity Search API is cheaper above 5,000. Exa charges $0.007 per standard request but supplies $10 in recurring monthly credit, while Perplexity charges $0.005 per raw Search request and publishes no recurring free allowance. Solving $0.007 × requests - $10 = $0.005 × requests gives the 5,000-request crossing. The separate $20 Exa signup credit is excluded because it does not recur. (Exa API pricing, Perplexity API pricing, retrieved 2026-07-26)

Search shape can move the verdict. Exa’s standard price includes up to ten results; each result above ten adds a published per-result charge, and requesting page contents or AI summaries adds another page meter. Perplexity’s raw Search endpoint publishes one request rate with no LLM token charge, and its reference exposes a result-count field without a separate public result-count surcharge in the reviewed pricing material. A request count is therefore comparable only when the requested output is held constant. (Exa API pricing, Perplexity Search reference, Perplexity API pricing, retrieved 2026-07-26)

Billing models side by side

Exa is pay as you go with recurring monthly credit. Standard Search costs $7 per 1,000 requests, Deep Search costs $12 per 1,000, and Deep-Reasoning Search costs $15 per 1,000. The selected search class replaces the other base classes; optional extra results, retrieved contents, and AI summaries are additive. Exa does not publish self-serve rate limits or failed-search billing on the reviewed pricing page. (Exa API pricing, retrieved 2026-07-26)

Perplexity Search API is also pay as you go, but the raw endpoint has one published rate of $5 per 1,000 requests and no token charge. It is commercially distinct from Sonar chat completions and Agent API tool calls, so their token and tool meters do not belong in this boundary. Perplexity also does not state failed-request billing or a Search-specific rate limit in the reviewed public pricing and reference pages. (Perplexity API pricing, Perplexity Search reference, retrieved 2026-07-26)

What each does that the other does not

Exa’s reviewed API pricing offers Deep Search and Deep-Reasoning Search as separately priced retrieval modes, and it can return page contents or AI summaries under explicit add-on meters. Perplexity’s raw Search endpoint is deliberately narrower: it returns ranked result records such as titles, URLs, snippets, and optional date fields without producing the prose completion associated with Sonar. A buyer who needs a deeper vendor-managed research operation can select an Exa deep mode; a buyer who needs Perplexity-generated prose must compare Sonar or Agent API instead of pretending raw Search provides it. (Exa API pricing, Perplexity Search reference, Perplexity API pricing, retrieved 2026-07-26)

Who should pick which

Pick Exa below 5,000 monthly standard searches when the recurring credit is usable, or when Exa’s deep modes, content retrieval, or summary add-ons replace application work. Pick Perplexity above 5,000 when the required output is the raw Search result array and the flat lower request rate remains the relevant meter. Recalculate Exa with the actual result count and page add-ons, and do not attach Sonar or Agent charges to a Perplexity raw Search call. (Exa API pricing, Perplexity API pricing, Perplexity Search reference, retrieved 2026-07-26)

The data-use edge also needs procurement review. Exa’s public terms restrict several forms of copying and distribution, while Perplexity’s Search-specific addendum and generic API FAQ conflict on query-data retention. Neither contract question changes the 5,000-request arithmetic, but either can disqualify the cheaper option for a particular application. (Exa terms, Perplexity Search addendum, Perplexity API FAQ, retrieved 2026-07-26)