Perplexity Chat

Yes, Perplexity Chat runs on the Sonar model family, which includes standard Sonar, Sonar Pro, and Sonar Reasoning Pro. While Perplexity Chat is the interface environment, Sonar is the underlying retrieval-augmented generation engine powering live search indexing and citation extraction.

The Generic variant routes directly through native infrastructure for the lowest possible retrieval latency and direct API handling, while the OpenRouter variant enables unified credential and budget management across multi-model architectures at the cost of slight routing overhead.

Choose the Generic endpoint if your workflow prioritizes minimal turnaround time and native search indexing, and choose the OpenRouter endpoint if you need centralized multi-provider orchestration under a single interface.

Standard Sonar delivers fast, low-cost lookups for concise facts; Sonar Pro expands the search context across dozens of live sources for multi-angle synthesis; and Sonar Reasoning Pro integrates Chain-of-Thought deduction for math, coding, and multi-step logic. Sonar Reasoning Pro displays internal <think> blocks and does not support image attachments.

Keep the temperature between 0.0 and 0.2 for maximum factual grounding and accurate citation matching. Higher values toward 1.0 can cause citation drift and hallucinated sources. Additionally, reasoning models prefix responses with <think> blocks, which require custom string handling if you are parsing output into structured formats.