Sonar puts Perplexity's cited real-time search behind an API at $1 per million tokens, undercutting the incumbents on price.
Research is the highest-volume, lowest-judgement task in a seller's week, and it is the one that most rewards automation. An API with citations attached means account research can run before a rep opens their laptop, with the sources still attached for checking.
An API — application programming interface — is how one piece of software calls another. Putting search behind an API means your own tooling can ask a question and get a cited answer back, without a person typing into a chat box.
Perplexity shipped two tiers. Sonar is the cheap fast one. Sonar Pro is built for complex, multi-step questions and costs substantially more.
The short version: the cited-web-search thing Perplexity does can now run inside your own systems, at a price low enough to run it on every account.
Perplexity launched the Sonar and Sonar Pro APIs on 21 January 2025, giving developers access to its real-time, citation-backed search.
Price: Sonar at $1 per million input and output tokens plus $5 per 1,000 searches. Sonar Pro at $3 per million input and $15 per million output tokens.
Benchmark: Sonar Pro reported an 85.8 F-score on the SimpleQA factuality benchmark.
Regions: Not stated.
Rate limits: Not stated in the launch coverage.
Verification: Sourced to TechCrunch's reporting rather than a Perplexity announcement post, so this item is marked Reported.
What it doesn't do: the product is called Sonar. It is not, and has never been, a 'Pro Search API' — a name that circulated widely and describes nothing Perplexity shipped. It also only reaches the open web: anything paywalled, private or inside your CRM stays invisible, so it complements your own data rather than replacing it.
SDR — You will not touch this directly, but you will feel it. If your ops team wires it up, pre-call briefs arrive already written and your job shifts from gathering to judging.
AE — The value is refresh rather than research. Standing account monitoring — what changed at these twenty accounts this week — is exactly what an API is for and exactly what nobody does manually.
Manager — Cheap automated research changes what 'prepared' means. If briefs are generated for everyone, preparation stops being a differentiator and the coaching bar moves to what the rep did with it.
RevOps — This is your build. Note the pricing has two axes — per-token and per-search — so cost modelling on a naive per-account estimate will be wrong.
Automated research is only as good as the question it answers. Fix the question first, by hand, on one real account.
Two minutes, once a week. What changed in AI, and what to run because of it.