Perplexity launched an API service called Sonar on Tuesday. This allows businesses and developers to incorporate the startup’s generative AI search tools into their own applications.
“Most generative AI capabilities today have answers derived solely from training data, but this limits their capabilities,” Perplexity wrote in a blog post. “To optimize facts and authority, APIs require real-time connectivity to the internet to get answers from trusted sources.”
First, Perplexity offers two tiers for developers to choose from. One is Sonar, a basic version that is cheap and fast, and the other is Sonar Pro, which is a more expensive version suitable for difficult questions. According to Perplexity, the Sonar API will allow businesses and developers to customize the sources that the AI search engine retrieves.
With the release of the API, Perplexity is making its AI search engine available in more places than just apps and websites. Perplexity says Zoom is already using Sonar to power its video conferencing platform’s AI assistant. Sonar enables Zoom’s AI chatbot to provide real-time answers based on information from web searches with citations, without the user ever leaving the video chat window.
Sonar could also give Perplexity another revenue stream, which could be especially important to the startup’s investors. Perplexity only offers a subscription service with unlimited access to its AI search engine and some additional features. However, the tech industry has reduced the price of accessing AI tools via APIs over the last year, and Perplexity claims to offer the cheapest AI search API on the market via Sonar.
The basic version of Sonar offers a cheaper and faster version of the company’s AI search tool. The basic version of Sonar has a flat fee and uses a lightweight model. It costs $5 for every 1,000 searches, plus $1 for every 750,000 words you input into your AI model (about 1 million input tokens), and an additional $1 for every 750,000 words your model spits out (about 1 million output tokens) .
The more expensive Sonar Pro provides more detailed answers and can handle more complex questions. This version performs multiple searches in addition to user prompts, which can make prices more unpredictable. Perplexity also states that this version provides twice the number of citations as the basic version of Sonar. Sonar Pro costs $5 for every 1,000 searches, plus $3 for every 750,000 words you input into your AI model (about 1 million input tokens), and $3 for every 750,000 words your model spits out (about 1 million output tokens). It costs $15.
Perplexity claims Sonar Pro outperformed leading models from Google, OpenAI, and Anthropic on SimpleQA, a benchmark that measures factual correctness in AI chatbot answers.
As we recently reported, Perplexity’s annual recurring revenue is between $5 million and $10 million. This seems pretty healthy for a startup of Perplexity’s size and age, but the startup is certainly looking for new ways to grow its revenue. The startup raised an additional $73.6 million in a funding round earlier this month, valuing the company at about $520 million.