Perplexity has launched the Sonar and Sonar Pro API, aimed at enhancing AI search capabilities with a focus on accuracy and real-time information. The Sonar Pro API, now available to developers, allows for the creation of generative search features that utilize web-wide research and trusted sources, including citations.
The Sonar API is designed to be lightweight, affordable, and fast, featuring customizable sources and optimized for speed, making it suitable for companies needing basic question-and-answer functionalities. In contrast, the Sonar Pro API supports complex, multi-step queries and offers double the citations per search, along with a larger context window for nuanced inquiries.
Numerous developers across various industries are leveraging Sonar for enhanced search experiences. For instance, Copy AI has improved research efficiency for sales teams, saving 8 hours per week per representative and increasing throughput by 20%. Doximity uses Sonar to provide doctors with quick, research-backed answers, emphasizing the importance of citations for accuracy in medicine. Zoom has integrated Sonar Pro into its AI Companion 2.0, enabling users to perform real-time searches during video calls.
In terms of factuality, Sonar Pro outperforms other models, achieving an F-score of 0.858 on the SimpleQA benchmark, while Sonar received an F-score of 0.773. This highlights Sonar Pro's capability to combine the summarization strengths of large language models with real-time data access, ensuring accurate responses to fact-seeking questions.
