Perplexity how to change model: pick one in the web prompt box

On Perplexity's website, you change the model with the dropdown at the bottom right of the prompt box. Choose Custom, pick a model, then set its reasoning and speed options. This guide to Perplexity how to change model follows Perplexity's Help Center article on choosing a preferred model, as of October 29, 2026.

One limit up front: that article gives steps for the web version only. It does not cover the iOS or Android apps or switching models inside an existing thread, so this guide does not either.

How to change the model on Perplexity

1. Open Perplexity on the web and go to the prompt box.
2. Open the model dropdown in the bottom right corner of the box.
3. Select Custom.
4. Choose a model from the model section, then pick the reasoning and speed options.

If you never choose a model, the dropdown reads Best. In that state Perplexity picks a model for each query itself.

Before you start

You need a Perplexity account on a plan that offers the model you want, and a browser. When you finish you will have a chosen model, with your reasoning and speed settings, set in the prompt box.

What "Best" means

Best is the default, and it is the only choice on the free plan. Perplexity's subscription models article says free users get Best and a limited number of Pro Searches, but cannot pick an advanced Search model by hand. Perplexity's article on the default model says the default is tuned for fast searches and browsing, and that Pro subscribers can use more advanced models from OpenAI and Anthropic for harder work such as coding and deeper reasoning.

If you are weighing a paid plan first, see our guide to whether Perplexity is free.

Which models you can pick

The list depends on your plan. As of October 27, 2026, the Help Center tables for Pro and Max include models from OpenAI, Google, Anthropic, Moonshot AI, Z.ai, xAI and NVIDIA. Several top models, such as GPT-6 Astra, Gemini 4 Argon and Claude Opus 5.5, are listed for Max only.

Enterprise Pro and Enterprise Max have similar lists. Administrators can restrict providers or models, and regional or data-retention settings can shorten the list further. Perplexity says models change over time, so the selector in your own account is the authoritative list.

Two things are worth knowing about the options. Thinking is a setting on supported models that gives them more time to reason. It is not a separate model. And Perplexity's modes, such as Deep Research, Learn Mode, Comet and Computer, are different from the Search mode. Computer and Comet have their own model selection.

Model Council runs several models at once

If you want answers from more than one model, Perplexity has a separate feature, Model Council. It sends one question to several models and merges the answers into one, showing where they agree and disagree.

On the web it is for Max and Enterprise Max subscribers. To turn it on, click Search below the search bar on the homepage, select Model Council, then enter your question. The search bar then shows "3 models". Click the model count to switch models on or off. Each model has its own Thinking toggle. Web and iOS have a monthly query cap.

If you don't see a model you want

The Help Center article says only that available models depend on your plan. Check the subscription models article for what your plan includes. API access is not part of a subscription and is billed separately, according to that same article.

The Help Center does not say how to switch models in an existing conversation, so check the dropdown in your own thread before assuming it works.

Perplexity's own list of models changes, and the dates in its tables show when it last checked. Look at the date on the Help Center page before you rely on a model name from this guide.