Sonar

Modellfamilie von Perplexity · verwendet von 1 Anwendung

KI-Modell

Sonar is Perplexity’s in-house large language model family, optimized for answer quality, user experience, and real-time factuality. The latest version is built on top of Llama 3.3 70B and further trained to enhance answer factuality and readability, especially for Perplexity’s default search mode. Sonar is designed to outperform models in its class (e.g., GPT-4o mini, Claude 3.5 Haiku) and closely match or exceed frontier models like GPT-4o and Claude 3.5 Sonnet in user satisfaction, while being faster and more cost-effectiveperplexity.ai+1. Sonar also powers Sonar Pro API, which delivers real-time, web-wide research and Q&A capabilities, with features like citations and customizable sources. It is widely used for generative search, deep research, and agentic workflowsperplexity.ai+1.