Skip to content
trending

Guides

Perplexity AI: Pricing, Models & Polish Test Results (2026)

Perplexity is an AI search engine that answers questions with numbered citations. Free version available, Pro costs $20/month. Our Polish-language test scored 9.5/10.

By 10 min read
Perplexity AI: Pricing, Models & Polish Test Results (2026)

👁 120 reads

Perplexity is an AI-powered search engine: you ask a question, it searches the web and answers with numbered citations to sources. The basic version is free, Pro costs $20/month ($99.99 in the Polish App Store), and Max $200/month ($999.99 in Poland). In our test of 10 questions about recent Polish events, the Sonar model scored 9.5 out of 10 and answered in under 2 seconds.

Status as of October 3, 2026

  • Plans: Free $0, Pro $20 USD, Max $200 USD, Education Pro, Enterprise Pro from $40 USD per person.
  • Agent: Perplexity Computer (launched February 25, 2026) is available in Pro and Max, paid with credits - 100 credits = $1 USD.
  • Browser: Comet has been free since October 2, 2025 on Windows and Mac, since November 2025 on Android, since March 2026 on iPhone.
  • New in September: Portable Computer on Windows with RTX cards, automations in Computer, model effort control (from Light to Ultra).
  • Our test: 10 questions about Polish current events via API - Sonar and Sonar Pro each scored 19/20 points, GPT-6.1 Sol 20/20, Gemini 3.8 Flash 17/20.

What is Perplexity

Perplexity works like a combination of search engine and chat. Instead of a list of ten links, you get a ready answer with a numbered source citation next to each statement. You can ask follow-up questions in the same thread, and the model remembers the conversation context.

The company was founded in August 2022 by Aravind Srinivas (CEO), Denis Yarats, Johnny Ho, and Andy Konwinski. The search engine launched on December 7, 2022. In February 2026, Perplexity abandoned advertising and focused exclusively on subscriptions, which we covered in Perplexity abandons ads. According to Wikipedia, the company’s valuation in early 2026 exceeded $21 billion USD.

Today Perplexity consists of four products under one account:

  • Search engine (perplexity.ai and apps) - questions, answers with sources, Research mode for longer reports.
  • Computer - an agent that breaks tasks into steps, uses dozens of models, and connects to applications (email, drive, Slack, GitHub). It also has a local version, Portable Computer, described in Portable Computer on Windows with RTX cards.
  • Comet - a Chromium-based browser with AI assistant. Details in the guide Perplexity Comet: browser with AI.
  • Sonar API - the same answers with sources for developers, paid per token and query.

How to get started step by step

  1. Go to perplexity.ai or download the app (Android, iOS, Windows, macOS). You can ask a few questions without logging in, but history and Spaces require an account - Google, Apple, or email is enough.
  2. Ask a question in a complete sentence. The more specific, the more accurate the sources.
  3. Check the citations. Click the number next to a statement and see if the page actually says that. How to do this quickly is described in Perplexity AI: how to search and verify sources.
  4. Ask follow-ups in the same thread instead of starting over - Perplexity retains context.
  5. For longer topics, turn on Research mode (formerly Deep Research). Example workflow: Perplexity Deep Research in practice.
  6. Collect repeatable projects in Spaces - with your own files and instructions. Guide: Perplexity Spaces step by step.

Modes and models

In the free plan, Perplexity automatically selects the model (“Best” option). In Pro and Max, you choose manually. The list changes every few weeks, and Perplexity’s help documentation states that the model selector in your account is the source of truth. According to a DataStudios compilation from September 17, 2026, it looked like this:

Model Provider Pro Max
Sonar 2 Perplexity yes yes
GPT-5.6 Terra OpenAI yes yes
GPT-5.6 Sol OpenAI no yes
Claude Sonnet 5 Anthropic yes yes
Claude Opus 5 Anthropic no yes
Gemini 3.1 Pro / 3.8 Flash Google yes yes
Grok 4.6 xAI yes yes

We found no confirmation that GPT-6 and Claude Opus 5.5 (released in late September) have reached the search engine selector yet. However, Opus 5.5 works as the main model for the Computer agent at standard effort level, and GPT-6 Sol and Luna support Agent API presets. Work modes:

  • Search / Pro Search - regular questions; Pro Search performs multiple searches and reads more pages.
  • Research - a report from dozens of sources in a few minutes, described in the guide Perplexity for research.
  • Computer - multi-step tasks with applications, paid with credits. Regular searches don’t consume credits.
  • Model Council (Max only) - the same question goes to multiple models at once, and a “chair” model combines the answers.

Pricing

Perplexity lists prices in dollars. When purchasing through the website, you pay in USD at your bank’s exchange rate.

Plan Price (USD) What you get
Free $0 unlimited search with sources, a few Pro queries daily, automatic model selection
Pro $20/month or $200/year model selection, Research, file uploads, Computer on credits, one-time 4,000 credits (valid 30 days)
Max $200/month or $2,000/year most expensive models, 10,000 credits monthly, Model Council, Brain, background assistants in Comet
Education Pro $10/month (sources vary) Pro for students and teachers after SheerID verification
Enterprise Pro $40 per person/month admin panel, no training on company data, 500 credits per person
Enterprise Max $325 per person/month Max features in enterprise version, 15,000 credits per person

Computer credits can be purchased in the app. One complex task can consume 100-300 credits, and failed tasks also consume them. We described the breakdown of plans and Pro drawbacks in Perplexity Pro: is it worth it and How much does Perplexity cost, and a comparison with other subscriptions in AI subscription prices.

Other legitimate ways to get Pro include the referral program (one month of Pro for each person referred, maximum 12) and the student plan.

Perplexity in Polish

The app interface is available in Polish (the App Store lists Polish among 52 languages), and the Google Play description is also translated. Answers come in the language of the question. Key point: when you ask in Polish, Perplexity searches Polish websites. In our test, the list of sources for questions about VAT on fuel or elections in Krakow consisted entirely of Polish sites - from rp.pl and tvn24.pl to bip.krakow.pl.

Weaker aspects in Polish:

  • Perplexity readily cites media instead of primary sources. For inflation, citations included not only the statistical office but also posts from x.com and salon24.pl.
  • When search results don’t contain an answer, the model says “I don’t know” instead of guessing - that’s a plus, but it can be overly cautious (question about the date of a central bank meeting).
  • Polish speech recognition in the Android app receives critical reviews in Google Play.

The full table of questions and answers is described in the test Perplexity in Polish: 10-question test.

Our test: 10 questions about Polish current events

On October 3, 2026, we asked ten questions via API (OpenRouter) about things that happened in Poland in recent days or changed recently: VAT and maximum fuel prices from October 3, September inflation, central bank rate, 2027 minimum wage, second round in Krakow, Poland-Romania match from October 2, pension indexation, euro exchange rate, time change. We tested in Polish. We verified each answer against official sources (statistical office, central bank, official gazette, electoral commission, central bank API). Full answers received 2 points, partial 1, incorrect or no answer 0.

Model (via API) Points (max 20) Official source in citations Average time Cost for 10 questions
Perplexity Sonar 19 5 out of 10 1.9 s $0.05
Perplexity Sonar Pro 19 5 out of 10 1.6 s $0.08
GPT-6.1 Sol with search 20 8 out of 10 22.1 s $0.29
Gemini 3.8 Flash with Google search 17 4 out of 10 5.1 s $0.26

Conclusions:

  • Perplexity is fastest and cheapest - answer in 2 seconds for half a cent. Sonar and Sonar Pro received identical sets of sources, so the difference only concerns the model writing the answer. In this test, the more expensive Sonar Pro added nothing.
  • GPT-6.1 Sol was the only one to get all 10 correct and most often reached primary sources (gov.pl, central bank, statistical office, official gazette), but we waited an average of 22 seconds for an answer.
  • Gemini 3.8 Flash once didn’t search at all - for the question about maximum Pb95 price, it considered the date to be in the future and stated there were no maximum prices, even though the minister’s announcement appeared the day before. For pensions, it attributed an expert’s forecast to the government.

This is an API test, not an app test - in the app, Perplexity Pro Search performs more searches than Sonar via API. The Sonar Pro Search model atlas page has parameters for the agent version.

Perplexity vs competitors

Tool Free version Paid plan Strength Weakness
Perplexity yes, unlimited basic searches $20/month citations with every statement, speed, model choice weaker writing and long-text work
ChatGPT with search yes Plus $20/month accuracy and primary sources in our test slower answers with sources
Gemini / Google AI Mode yes Google AI Pro ~$20/month Google index, local data in our test (Gemini via API) once skipped searching, links through redirects
Microsoft Copilot yes Microsoft 365 Premium ~$20/month Windows and Office integration wasn’t in our test, so we don’t evaluate accuracy

Longer comparisons: Perplexity vs ChatGPT Search vs Google, AI Mode in Google, Microsoft Copilot, and ChatGPT from the ground up. If you want Perplexity alongside other models in one app, check out Poe - it has Perplexity Sonar bots.

Who is Perplexity for

  • Anyone looking for facts - prices, regulations, dates, results. The free version is enough for a dozen questions a day.
  • Journalists and analysts - citations shorten verification. We described this in Perplexity for journalists.
  • Developers - Sonar API costs $1 per million tokens plus $5-12 per thousand queries. Guide: Perplexity API from scratch.
  • Not for people who mainly write long texts, code, or generate images - ChatGPT, Claude, and Gemini work better there.

All Perplexity articles on this site

Here we’ve collected everything we’ve written about Perplexity, grouped by topic.

Pricing, plans and reviews

Guides

Comet, Computer and new features

Company, advertising and disputes

  • Perplexity abandons ads (September 2026)
  • Perplexity says goodbye to ads (March 2026)
  • Perplexity focuses on ads - earlier turn
  • Licensing agreements: Elsevier, PLS
  • Controversies around unlicensed content
  • Perplexity’s impact on journalism
  • Perplexity in education and research
  • History of Perplexity features

Older guides (archive)

  • Perplexity - complete guide
  • Perplexity in Polish - practical guide (2025)
  • Guide to AI search engine (2025)
  • Guide from 2024
  • The future of browsing
  • How Perplexity changes search
  • Perplexity and access to information

Frequently asked questions

Is Perplexity free?

Yes. The Free plan gives unlimited basic search with sources and a few Pro queries daily, with automatic model selection. The Comet browser is also free. You only pay for model selection, Research at larger scale, and the Computer agent.

How much does Perplexity Pro cost?

Perplexity Pro costs $20 monthly or $200 annually. Max costs $200 monthly or $2,000 annually. When purchasing through the website, you pay in USD at your bank’s exchange rate.

Does Perplexity work in Polish?

Yes. The app has a Polish interface, and it answers Polish questions in Polish and searches Polish websites. In our test of 10 questions about current Polish events, Sonar scored 19 out of 20 points.

Perplexity is faster and shows a citation with every statement. ChatGPT with search got all questions right in our test and more often cited primary sources, but answered more than ten times slower. For quick fact-checking I choose Perplexity, for analysis with writing ChatGPT.

What models are in Perplexity?

In Pro: Sonar 2, GPT-5.6 Terra, Claude Sonnet 5, Gemini 3.1 Pro and 3.8 Flash, Grok 4.6. Max adds GPT-5.6 Sol and Claude Opus 5. The Computer agent also uses Claude Opus 5.5. The list changes frequently; the current one is visible in the model selector.

Does Perplexity hallucinate?

Less often than chat without search, but it happens that a citation leads to a page that says something different. We always click on the source for numbers and dates. Defense methods are described in the article about AI hallucinations.

What is Perplexity Comet?

Comet is Perplexity’s free Chromium-based browser with an AI assistant that summarizes pages, compares tabs, and performs tasks in the browser. It works on Windows, macOS, Android, and iOS.

How to get Perplexity Pro for free?

Legally through partner promotions, the referral program - one month of Pro for each person referred, up to 12 months - or the education offer. Before the free period ends, turn off automatic renewal.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Sources and verification date

We verified all data on October 3, 2026. We conducted the API test the same day through OpenRouter, verifying reference answers in primary sources.

Kinetyra

// read next

More on this from QUANTSLANT

Piotr Olszewski

Editor and publisher · Quantslant

I edit and publish Quantslant. AI tools help with research, drafts and translation; choosing topics, running the tests, checking facts and the decision to publish are mine, and so is the responsibility for what appears here.

read nextKiro (AWS): Pricing, Models, Specs & CLI Guideai toolsKiro (AWS): Pricing, Models, Specs & CLI GuideAWS Kiro IDE with AI agent: spec-driven development, credit pricing, Claude 5.5 & GPT-5.6 models, hooks, steering, and Kiro CLI. Free plan with 50 credits.

Most read

  1. 1Best AI Video Generator Apps for Your Phone (2026)
  2. 2Claude Opus 5.5 Motion Graphics: 60s Showreel from One Prompt
  3. 3Wan 2.2 on RTX 4090: timings, VRAM and four test clips
// site map

This page in other languages6

Quantslant in numbers 1,423posts in archive22news from the last 7 days18video models in the observatory
🎬 AI Video Benchmarks: Higgsfield DoP vs OpenAI Sora • Higgsfield vs Runway Gen-3 • AI Video Credit Calculator
× ‹ enlarge ›