When I switch between models in Perplexity, I’m not sure whether I’m changing only the model that writes the response or also changing how sources are found and selected. That distinction matters when I’m comparing answers: a different summary could come from different reasoning, a different set of search results, or both.

Has anyone tested this with the same question and the same search settings? I’d be interested in whether the selected model affects the search and retrieval pipeline, or mainly the final synthesis of the sources Perplexity has gathered. If the behavior varies by mode or plan, that would be useful context too.