Sonar Large Online

Sonar Large Online is a text generation model developed by Perplexity, built on the Llama 3.1 architecture and identified internally as llama-3.1-sonar-large-128k-online. It features a context window of 127,072 tokens and a maximum response size of 28,000 tokens, making it suitable for tasks that require processing or generating substantial amounts of text in a single session. The model is part of Perplexity's Sonar family, which was designed to improve on earlier Sonar versions in cost-efficiency, speed, and overall performance.

What distinguishes Sonar Large Online from offline language models is its live internet access, which allows it to retrieve and synthesize current information at inference time rather than relying solely on a static training dataset. It supports optional citation and image return through configurable inputs, making it useful for research assistance, fact-checking, and question answering where source attribution matters. The model has been marked as deprecated on MindStudio, meaning it has been succeeded by newer versions in the Sonar lineup. Developers working on applications that require up-to-date information retrieval with traceable sources would find this model's architecture particularly relevant.