Perplexity

Perplexity

Provider

Perplexity

Price

$0-325/user/mo

Platforms

Web, iOS, Android

Launched

Dec 2022

Updated June 2026

Perplexity is an AI answer engine: you ask a question, and it returns a direct, written answer with numbered citations to the web pages it drew from, rather than a list of links. It is the flagship product of Perplexity AI, the San Francisco company founded in August 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho and Andy Konwinski, and it launched its public answer engine on 7 December 2022.

What sets Perplexity apart from a general assistant like ChatGPT is its starting point. Perplexity is built around search and sourcing first: every answer is grounded in live web results and shows its citations inline, which is why it is most often used as a research tool and a replacement for traditional search. In 2026 it has grown into a wider platform, with a model picker that lets paying users run GPT-5.2, Claude, Gemini and Grok alongside its own Sonar models, a Deep Research mode, and Comet, an agentic AI browser that can carry out tasks on the web for you.

This guide covers the whole picture: the free answer engine, the paid Pro, Max and Enterprise tiers, the multi-model engine room, the Comet browser, and where Perplexity genuinely leads versus where it falls short.

Quick stats

Answer engine, not chatbot. Perplexity’s defining feature is that every answer carries inline citations to its web sources. It is designed to be checked, which is why it reads less like a conversation and more like a researched brief.

Pricing breakdown

Perplexity runs a free tier plus four paid tiers, and a separate $5/month publisher tier (Comet Plus) that is bundled into Pro and Max. All prices are in USD.

Free — $0

The free plan is a complete answer engine, not a crippled demo. It includes:

• Unlimited Quick Search on Perplexity’s default in-house Sonar model
• A limited number of Pro Searches per day — the slower, multi-step, more source-rich mode
• The Comet AI browser and Comet Assistant, which Perplexity made free worldwide around 2 October 2025
• File uploads, basic Spaces, the Discover feed, and the mobile and desktop apps

The free tier is the main reason Perplexity is so widely used: most casual questions never hit the daily Pro Search limit.

Pro — $20/month

Pro is the tier most individuals choose, at $20/month or $200/year. It adds:

• 300+ Pro Searches per day rather than a handful
• The model picker — choose GPT-5.2, Claude Opus 4.6, Gemini 3 Pro, Grok 4 or the Sonar family per thread
• Deep Research for long, multi-source reports
• Labs, which can build small apps, dashboards and analyses from a prompt
• Larger file uploads, fuller Spaces, and bundled Comet Plus

Max — $200/month

Max is the power-user tier at $200/month or $2,000/year. It adds, on top of Pro:

• The highest usage limits across search and research
• Earliest and priority access to top frontier reasoning models
• Model Council — sends one query to three frontier models and synthesises their answers (Max-exclusive, launched 5 February 2026)
• The Comet background assistant, which can run longer tasks for you
• Priority support

Enterprise Pro — $40/user/month

The team and business tier, at $40/user/month or $400/year. It adds organisation features to Pro:

• SSO, admin controls and user management
• Internal Knowledge — search across connected company files alongside the web
• SOC 2 Type II compliance, with data never used for training and enterprise-grade encryption

Enterprise Max — $325/user/month

The top enterprise tier, at $325/user/month or $3,250/year, aimed at organisations running heavy research and agentic workloads. It adds the highest model and performance tier — Max-level reasoning models, Model Council and advanced agentic features — to Enterprise Pro.

Students and promotions. Verified students can access discounted Pro (aggregators cite an Education Pro rate around $10/month), and Perplexity has repeatedly run campaigns and partner promotions offering free Pro for a period. Check the current student and partner offers before paying full price.

Sonar API

Developers can call Perplexity’s search-grounded models directly through the Sonar API, billed per use rather than by subscription. The API carries a zero-data-retention policy: requests are not stored and are not used for training.

Models powering Perplexity

Perplexity’s most distinctive technical choice is that it is multi-model. It runs its own search-tuned models by default, and lets paying users swap in the major frontier models per question.

Sonar — the in-house default

Sonar is Perplexity’s own model family, fine-tuned on Meta’s Llama (3.3 70B base) and trained specifically for fast, factual, citation-linked search. Perplexity reports Sonar scores an F-score of 0.773 on the SimpleQA factuality benchmark and, in its own A/B testing, matches or beats frontier models for user satisfaction on search tasks while being much cheaper to run. Sonar is not a from-scratch foundation model; it is a search-optimised offshoot. Variants include Sonar (default), Sonar Pro and Sonar Reasoning Pro.

The frontier-model picker

On Pro and Max, you can choose the model per thread. The picker covers the current major frontier models — GPT-5.2, Claude Opus 4.6, Gemini 3 Pro and Grok 4 — and is updated as new versions ship, so the exact list moves over time. Max subscribers get the earliest and priority access to the top reasoning models. Two automatic modes sit alongside the manual picker: Best, which auto-selects the most suitable model, and Research, which composes a mix of models for a deep, source-rich report.

Model Council (Max)

Model Council, launched 5 February 2026, is a Max-exclusive feature that sends a single query to three frontier models at once — at launch, Claude Opus 4.6, GPT-5.2 and Gemini 3 Pro — and produces a synthesis that highlights where they agree, disagree and add unique insight. It is Perplexity’s answer to the question of which model to trust: run several and compare.

Features deep dive

The answer engine

The core product returns a written answer with numbered inline citations to its web sources, plus suggested follow-up questions. Pro Search runs a slower, multi-step process across more sources; Quick Search answers fast on Sonar. This sourcing-first design is the whole point: answers are built to be verified.

Deep Research

Perplexity’s Deep Research mode runs an extended, multi-source investigation and returns a long, structured report with citations — the direct equivalent of ChatGPT’s and Gemini’s deep-research features. It is aimed at questions that would otherwise take a person an afternoon of reading.

Comet — the agentic browser

Comet is Perplexity’s Chromium-based AI browser. It first launched on desktop in July 2025 to $200/month Max subscribers, went free worldwide around 2 October 2025, reached Android in November 2025 and iOS in March 2026. Its centrepiece is the Comet Assistant, an agent that can read the pages you are on and carry out tasks across multiple tabs — finding flights, organising a job search, filling forms, comparing products. Perplexity reimagined the Comet Assistant on 6 November 2025, reporting it performs 23% better than its predecessor on internal multi-step tests, and gave Max users a background assistant that can work on longer jobs. Comet asks permission before taking browser actions on your behalf.

Labs, Spaces, Pages and Discover

Labs turns a prompt into a small working artefact — an app, dashboard, spreadsheet or simple analysis. Spaces are collaborative collections that bundle threads and uploaded files around a topic, with their files retained rather than auto-deleted. Pages turns a research thread into a shareable, formatted document. Discover is a curated feed of current stories, each expandable into a full Perplexity answer.

Assistant, Shopping, Finance and Voice

The mobile Perplexity Assistant is an agentic helper that can act across apps on your phone. Shopping surfaces product research and, in supported regions, a “Buy with Pro” checkout, while Finance and Travel views package live data into answerable dashboards. A voice assistant lets you ask questions hands-free.

Strengths

Cited, verifiable answers by default — Every response shows its sources inline, which makes Perplexity the natural fit for research and fact-finding and easy to check.

Genuine multi-model access — One subscription gives you Sonar plus GPT-5.2, Claude, Gemini 3 Pro and Grok 4, switchable per question, and Model Council to compare them. No mainstream rival offers this breadth under one roof.

A capable, generous free tier — Unlimited Quick Search, a daily allowance of Pro Search, and the full Comet browser at no cost make it one of the most usable free AI products.

Comet and agentic browsing — A free, cross-platform agentic browser is a real differentiator; the Comet Assistant can complete multi-step web tasks rather than just answering.

Search-first speed and focus — Sonar is tuned for fast, factual answers, so routine questions return quickly and stay on topic.

Limitations

Best for research, weaker for freeform creativity — Perplexity’s sourcing-first design serves questions and research far better than long, open-ended creative writing, where a general assistant is usually stronger.

Citation quality still needs checking — Grounded answers reduce, but do not eliminate, errors; sources can be thin or mismatched, so claims still warrant a glance at the citations.

Publisher and legal disputes — Perplexity faces copyright litigation from Dow Jones and the New York Post, a New York Times suit, and an Amazon injunction (March 2026) blocking Comet’s agent from Amazon — ongoing disputes that could affect content access and agentic features.

Tier value gap — The jump from Pro ($20) to Max ($200) is large; Max only pays off for genuinely heavy users who need Model Council, the highest limits and earliest model access.

No single bundled model — The model picker is a strength for power users but adds choice and complexity for people who just want one good default.

Who should use Perplexity?

Great for:

• Researchers, analysts and students — cited answers and Deep Research make sourcing fast and checkable
• People who want a better search engine — direct answers with references instead of ten blue links
• Power users who compare models — Sonar plus GPT-5.2, Claude, Gemini and Grok, with Model Council on Max
• Agentic-browsing early adopters — Comet can carry out web tasks, free across platforms

Look elsewhere if:

• You mainly want long creative writing — ChatGPT or Claude are stronger for freeform prose
• You want one simple bundled model — Perplexity’s picker is built for choice
• You need a coding-first tool — a dedicated coding assistant will serve you better

Historical timeline

Privacy and data handling

Perplexity’s data treatment differs sharply by plan.

Enterprise Pro and the Sonar API: Perplexity states that Enterprise Pro data is not used to train its models, and the Sonar API operates a zero-data-retention policy. Perplexity is SOC 2 Type II compliant and references GDPR and PCI DSS alignment. Files attached to threads are deleted after seven days; files in Spaces are retained and encrypted at rest and in transit.

Consumer plans (Free, Pro, Max): Consumer interactions can be used to improve the service, with an AI Data Retention opt-out in settings. Incognito threads are not saved to history and expire within 24 hours. Check the Perplexity Trust Center and privacy policy for current details.

A note of context: Perplexity’s web-access methods are contested. Several publishers allege their content is reproduced too closely, and there is active litigation (see Limitations). This is relevant to how the product sources answers, even though it is separate from how it handles your personal data.

Perplexity vs alternatives

Quick comparison

Where Perplexity wins

• Cited, sourcing-first answers built to be checked
• Multi-model choice in one subscription, plus Model Council on Max
• A free agentic browser (Comet) across desktop and mobile

Where competitors win

• ChatGPT — broader general-purpose use, stronger creative writing, larger ecosystem
• Gemini — very large context windows and deep Google Workspace integration
• Microsoft Copilot — grounded work assistance inside Microsoft 365 and Windows

For a fuller head-to-head on research specifically, see our best AI for research guide.

Getting started

Free answer engine

1. Go to perplexity.ai or download the app for iOS or Android
2. Ask a question — Quick Search answers immediately, with citations
3. Open a citation to verify, or ask a follow-up

Comet browser

Download Comet for desktop, Android or iOS, sign in, and use the Comet Assistant to ask questions about the page you are on or to run a multi-step task.

Pro, Max and Enterprise

Upgrade in account settings for the model picker, higher limits and Deep Research. Teams buy Enterprise Pro or Enterprise Max with admin controls and internal-file search.

Frequently asked questions

Is Perplexity free?

Yes. The free plan includes unlimited Quick Search on the Sonar model, a daily allowance of Pro Search, and the full Comet browser. Pro ($20/month) adds the frontier-model picker, far higher Pro Search limits and Deep Research.

What is the difference between Perplexity and ChatGPT?

Perplexity is an answer engine built around web search, and every answer shows inline citations. ChatGPT is a general-purpose assistant that is stronger for open-ended and creative tasks. Perplexity is usually the better fit for research and fact-finding; ChatGPT for broad, conversational use.

Which AI models does Perplexity use?

Its own search-tuned Sonar models by default, plus a per-thread picker on Pro and Max covering GPT-5.2, Claude (Opus 4.6 and Sonnet), Gemini 3 Pro and Grok 4. The exact list updates as new models ship.

How much does Perplexity Pro cost?

$20/month, or $200/year billed annually. Max is $200/month; Enterprise Pro is $40/user/month and Enterprise Max is $325/user/month.

Is the Comet browser free?

Yes. Comet went free worldwide around 2 October 2025 and is available on desktop, Android and iOS. Max subscribers additionally get a background assistant for longer tasks.

Does Perplexity train on my data?

Consumer interactions can be used to improve the service, with an opt-out in settings; Incognito threads are not saved. Enterprise Pro data is not used for training, and the Sonar API has zero data retention.

What is Model Council?

A Max-exclusive feature (launched 5 February 2026) that sends one query to three frontier models at once and synthesises their answers, highlighting agreement, disagreement and unique insight.

Verdict

Perplexity has carved out a clear identity: the AI that answers your question and shows its sources. That sourcing-first design makes it the default tool for research, fact-finding and anyone who wants to replace a search engine with something that gives a direct, checkable answer. Its multi-model approach — Sonar by default, with GPT-5.2, Claude, Gemini and Grok a click away, and Model Council on Max — is a genuine differentiator no mainstream rival matches.

Free tier: Unusually capable, with unlimited Quick Search, a daily Pro Search allowance and the full Comet browser — worth using on its own.

Pro ($20/month): The sweet spot for individuals, unlocking the model picker, high Pro Search limits and Deep Research.

Max ($200/month): Only for heavy users who need Model Council, the highest limits and earliest model access.

The honest caveats: Perplexity is a research and search tool first, so a general assistant is better for long creative work; grounded answers still need their citations checked; and the company’s web-access methods face active litigation from major publishers and Amazon. But for the core job of getting a direct, sourced answer fast — and, increasingly, having an agent act on it through Comet — Perplexity is among the most useful AI products available.

Official links