Perplexity Sonar
Search-grounded answer API: one OpenAI-compatible call returns cited prose from live web search.
About
Sonar folds web search and model reasoning into a single API call. You send a question to the OpenAI-compatible endpoint at api.perplexity.ai and get back a grounded prose answer with inline citations, instead of raw results you have to synthesize yourself. The lineup spans sonar, sonar-pro, sonar-reasoning-pro, and sonar-deep-research for different depth needs, with configurable search context size and domain filters.
A separate Search API returns classic ranked results (titles, URLs, snippets, dates) for agent loops that need to inspect sources themselves, and an Agent API wires Perplexity's web tools into third-party models. Migration from any OpenAI client is famously a two-line change: swap the base URL and the model name.
Pricing is the thing to model carefully: Sonar bills tokens plus a per-request search fee, so costs stack at volume, and there is no free tier. Independent 2026 comparisons put it mid-pack on price with the shortest path from question to shippable cited answer. Also worth tracking: Reddit's 2025 scraping suit names Perplexity, a consideration for long-lived dependencies.
What it can do
- Single-call grounded answers with inline citations
- Model lineup: sonar, sonar-pro, sonar-reasoning-pro, sonar-deep-research
- OpenAI-compatible Chat Completions endpoint
- Raw Search API plus Agent API for third-party models
- Domain filters and configurable search context depth
Setup
Create an API key at perplexity.ai; point any OpenAI-compatible client at https://api.perplexity.ai with a sonar model name.
Prerequisites
A Perplexity API key and billing set up; no free tier.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: OpenAI-compatible endpoint and model lineup verified against official docs; pricing ($5/1k search, token+request fees) corroborated by multiple independent 2026 API comparisons; No free tier; token fees stack on per-request search fees at volume. Manual review: Reddit's 2025 scraping suit names Perplexity -- worth tracking for long-lived dependencies; cited-answer quality claims are provider-framed. License: Proprietary (API-only).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 3.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · OpenAI-compatible endpoint and model lineup verified against official docs; pricing ($5/1k search, token+request fees) corroborated by multiple independent 2026 API comparisons
- · No free tier; token fees stack on per-request search fees at volume
- · manual review: Reddit's 2025 scraping suit names Perplexity -- worth tracking for long-lived dependencies; cited-answer quality claims are provider-framed
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.