Sonar is a streamlined answer engine model from Perplexity designed to pair natural language generation with live web search, producing responses that read like concise, source-backed summaries rather than open-ended chat. The model accepts both text and image inputs and returns text output, making it well suited to question answering, quick research tasks, and conversational lookups where freshness and verifiability matter more than long-form reasoning. Its design intent centers on speed and groundedness: Sonar is built to retrieve relevant information from the open web and weave it into a compact, readable reply, so it behaves less like a general creative writer and more like a research assistant optimized for clarity and citations in everyday use.

Within the broader Sonar family, this variant represents the lighter, lower-latency tier intended for high-volume, cost-efficient deployment while still leveraging Perplexity's search-augmented generation stack. In practice that translates to strong fit for retrieval-heavy workflows such as news summarization, fact lookups, product comparisons, and multimodal prompts where an attached image needs to be interpreted alongside live web context. Because the model routes queries through real-time search infrastructure, its answers tend to reflect current information rather than a static training corpus, which makes it a natural choice for applications that need up-to-date grounding, lightweight agentic steps, or quick summarization tasks where verifiable, citation-friendly output is the primary goal.