Perplexity Sonar

Perplexity Sonar is positioned as a lightweight, affordable, and fast question-and-answer model built for companies that want to weave real-time, citation-backed search into their products and internal tools. According to its public listing, it ships with built-in citations and the ability to customize which sources it draws from, letting teams tune retrieval to their own domains instead of relying on a generic web crawl. The design emphasis on speed and simplicity makes it a practical fit for chatbots, in-app help, and developer workflows where low-latency answers with traceable references matter more than open-ended generation.

From a deployment standpoint, the model is offered through a single upstream provider with direct request forwarding, and third-party routing catalogs report strong uptime along with a P50 latency near 1.6 seconds and throughput around 79 tokens per second on that route. A 127,000-token context window leaves ample room for long documents, multi-source synthesis, and retrieval-augmented prompts, while flat per-million-token pricing on both input and output keeps cost forecasting straightforward for high-volume Q&A workloads. Independent developer coverage predating the public release confirms that Sonar was being adopted as an enterprise search API for IDE and workflow integrations, reinforcing its reputation as a focused retrieval-and-answer model rather than a general-purpose chat model.