Perplexity

Rating reflects Perplexity's documented plans and capabilities as of September 2026 plus our own sessions. Some vendor links on this site are affiliate links; they never affect a rating. See our disclosure and method.

What it does

Type a question and Perplexity searches the web, reads the results, and writes an answer with numbered citations you can click. That loop is the whole product, and the rest of the app is built around making it deeper. Four search modes sit in the search box: Best, which picks a model for you and has no quota; Pro Search, which runs several searches and reasons across them; Reasoning, for questions that need multiple steps; and Research, which runs for minutes, reads hundreds of pages and returns a structured report. Research picks its own models and will not let you choose one. You can upload PDFs, spreadsheets, images, audio and video and ask about them, with a weekly upload allowance that is higher on Max than on Pro.

The model picker is the feature that separates Perplexity from every other card in this suit. Pro subscribers choose which model answers: Perplexity's own Sonar, or third-party models from OpenAI, Anthropic and Google. Perplexity's help center listed GPT-5.2, Claude Sonnet 4.6 and Gemini 3.1 Pro on the picker in its September update, with Grok 4 and Anthropic's thinking models for harder questions; the company adds frontier models as they ship, and its April changelog already had Claude Opus 4.7 and GPT-5.4 as orchestrator options. Model Council, a Max-era feature that has spread to Computer, asks two to eight models the same question and shows where they agree and disagree.

Computer is the newer half of the product: an agent that runs multi-step work in the background, from market research across dozens of sites to building spreadsheets, dashboards and small apps, with connectors to Gmail, Outlook, GitHub, Slack, Notion and Salesforce. It runs on credits, where 100 credits equal $1. Max includes 10,000 credits a month; Pro has no monthly allocation but comes with a one-time 4,000-credit bonus, which is the "$40 free Computer credits" on the Pro sign-up page. On September 17 Perplexity added an effort slider, from Light to Ultra, that picks the model and reasoning depth for each task and spends credits accordingly.

Comet, Perplexity's Chromium browser, launched in July 2025 as a Max perk, went free for everyone that October, and reached iOS in March 2026. An assistant panel reads the page or any tab you mention with @, summarizes, drafts replies and extracts dates and names. Max subscribers get Max Assistant, a heavier reasoning mode with the highest weekly limit on browser-agent queries.

Where it wins

• Cited answers. Every claim has a numbered source you can open. For fact-finding, buying decisions and "what changed this week" questions, that is faster and more checkable than any chatbot, including ChatGPT's search mode.
• Model choice. One $20 plan gives you GPT, Claude and Gemini models side by side, and Model Council lets you make them argue. If you cannot decide which assistant to pay for, this is the hedge.
• Research mode. Independent comparisons through 2026 put Perplexity's reports at three to ten minutes with 40 to 100 sources, the fastest of the major deep-research modes. It is the mode we reach for when a topic has a dozen competing vendor pages.
• Comet for free. A page-aware assistant in the browser, at no cost, on every platform. Researchers and writers who live in tabs will get real work out of it.
• Files and connectors. Uploads cover audio and video with automatic transcription, Spaces hold shared files for a team, and Computer reaches into your email and tools when a task needs them.

Where it loses

• Prose and reasoning. It is an answer engine, not a writing partner. For drafting, editing, tone and long conversations, ChatGPT and Claude are better, and Perplexity's own models are a step behind the frontier.
• Credits creep. Pro searches never touch credits, but Computer does, and in September the browser-control feature that Pro users had in Comet moved behind Computer credits. Read the usage page before you assume a feature is in your plan.
• Training default and third parties. Consumer plans are opted in to training unless you switch it off, and your queries go to whichever third-party model you picked under that provider's terms. Comet's agent has also been the subject of published prompt-injection exploits; keep it away from your bank.
• Max is expensive for what most people need. $200 buys higher limits, Computer credits and early features. Unless you run Computer daily, Pro is the ceiling.
• Citation quality varies. Sources are shown, but they are not always the best sources, and a confident summary of a weak page reads exactly like a confident summary of a good one. Click through.

Plans and pricing, September 2026

Enterprise Pro starts at $40 a seat a month and Enterprise Max lists at $325 a seat, with organization-level file controls, admin settings and no training on company data. Developers use the Agent API, which replaced the Sonar model lineup in August 2026 and is billed separately.

Score breakdown

Headline rating weights Capability 30%, Ease of use 20%, Value 20%, Trust 15% and Ecosystem 15%, the same weights we use for every card.

Alternatives

• ChatGPT if you want one assistant that also writes, codes and talks, with search built in.
• Gemini if you want Deep Research inside Google's apps with the storage bundle attached.
• Claude if the research is a step toward a long document you will actually write.

Quick answers

Is Perplexity free?

Yes. The free Standard plan gives you unlimited basic searches with citations, a very limited number of Pro Searches, limited Research mode and basic file uploads, with Perplexity choosing the model for you. The Comet browser is also free. Pro is $20 a month, or $17 a month billed annually.

What is the difference between Perplexity Pro and Max?

Pro at $20 unlocks extended Pro Search, extended Research, more file uploads and the model picker with models from OpenAI, Anthropic and Google. Max at $200 adds the highest model access, 10,000 Computer credits a month, Max Assistant in Comet, the Brain memory preview and priority support. Most people should stop at Pro.

Does Perplexity train on my searches?

By default, yes, on Free, Pro and Max. You can opt out under the AI data retention setting in your account, and Enterprise plans are excluded from training. Perplexity also passes queries to third-party models such as GPT and Claude, under those providers' terms.

More cards

• 
  Tool cards

  ChatGPT

  The model-first assistant, rated and priced.

• 
  Tool cards

  Gemini

  The Google-native assistant and its bundle.

• 
  Prompt packs

  Learn anything faster

  Prompts that turn any assistant into a patient tutor.