Perplexity on Tuesday launched an API service referred to as Sonar, permitting enterprises and builders to construct the startup’s generative AI search instruments into their very own functions.
“While most generative AI options immediately have solutions knowledgeable solely by coaching knowledge, this limits their capabilities,” Perplexity wrote in a weblog put up. “To optimize for factuality and authority, APIs require a real-time connection to the Internet, with solutions knowledgeable by trusted sources.”
To begin, Perplexity is providing two tiers that builders can select from: a base model that’s cheaper and quicker, Sonar, and a pricier model that’s higher for powerful questions, Sonar Pro. Perplexity says the Sonar API additionally offers enterprises and builders the flexibility to customise the sources its AI search engine pulls from.
With the launch of its API, Perplexity is making its AI search engine out there in additional locations than simply its app and web site. Perplexity says that Zoom, amongst different firms, is already utilizing Sonar to energy an AI assistant for its video conferencing platform. Sonar is permitting Zoom’s AI chatbot to present real-time solutions, knowledgeable by net searches with citations, with out requiring customers to depart the video chat window.
Sonar might additionally give Perplexity one other income, which could possibly be explicit vital to the startup’s buyers. Perplexity solely presents a subscription service for limitless entry to its AI search engine and a few further options. However, the tech trade has slashed costs to entry AI instruments through APIs within the final 12 months, and Perplexity claims to offer the most affordable AI search API in the marketplace through Sonar.
The base model of Sonar presents a less expensive and faster model of the corporate’s AI search instruments. Sonar’s base model has flat pricing and makes use of a light-weight mannequin. It prices $5 for each 1,000 searches, plus $1 for each 750,000 phrases you sort into the AI mannequin (roughly 1 million enter tokens), and one other $1 for each 750,000 phrases the mannequin spits out (roughly 1 million output tokens).
The pricier Sonar Pro offers more-detailed solutions and is able to dealing with more-complex questions. This model will run a number of searches on prime of a consumer immediate, that means the pricing could possibly be extra unpredictable. Perplexity additionally says this model presents twice as many citations as the bottom model of Sonar. Sonar Pro prices $5 for each 1,000 searches, plus $3 for each 750,000 phrases you sort into the AI mannequin (roughly 1 million enter tokens), and $15 for each 750,000 phrases the mannequin spits out (roughly 1 million output tokens).
Perplexity claims Sonar Pro outperformed main fashions from Google, OpenAI, and Anthropic on a benchmark that measures factual correctness in AI chatbot solutions, SimpleQA.
As we lately reported, Perplexity’s annual recurring income is someplace between $5 million and $10 million. This appears pretty wholesome for a startup of Perplexity’s dimension and age, however the startup is definitely in search of new methods to develop its income. The startup raised an extra $73.6 million in a funding spherical earlier this month, valuing the corporate round $520 million.
