Perplexity AI
Perplexity is an AI-powered search engine that gives you cited, synthesized answers instead of a list of links. You ask a question in natural language and Perplexity retrieves real-time web results, reads the sources, and writes a response with inline citations — so you can verify every claim. It is positioned as a direct alternative to Google for research and knowledge queries.
How It Works
Every Perplexity query triggers a live web search. The model reads the retrieved pages and synthesizes an answer from them — it is always grounded in current web content, not training data alone. This makes it fundamentally different from asking a chatbot: the response reflects what is on the web right now, with sources you can click through.
Key Features
Pro Search
Orchestrates multiple sequential web searches, fetches content from multiple URLs, and reasons through results before answering. Significantly deeper than a single-pass search. Available on paid plans with a daily limit on Free.
Deep Research
Extended multi-step analysis mode — plans a research strategy, evaluates sources, and synthesizes a long-form report across many sources before responding. Best for complex, multi-faceted questions.
Pages
Generates structured, report-style documents from your query — aggregated, cited, and formatted. One-click research brief that you can share as a link or export.
Spaces
Shared collections of search threads for teams. Members can view research threads, add files, and build on each other's queries. Available on Enterprise plans.
Model Council (Feb 2026)
Compare outputs from multiple LLMs — GPT-4o, Claude, Gemini — simultaneously on the same query. Useful for seeing different perspectives or validating answers.
Model Selection (Pro)
Pro users can switch the backend model: default Sonar (Llama-based), GPT-4o, Claude 4.5/4.6, or Gemini 2.5 Pro. Each gives different answer styles for the same search.
Perplexity Computer (Max — Feb 2026)
An autonomous AI agent exclusive to Max subscribers ($200/mo). Breaks complex, multi-step tasks into sub-tasks and orchestrates up to 19 specialised AI models — including Claude Opus, GPT-5, Gemini, and Grok — running in a secure cloud sandbox. Tasks can run for hours or days without supervision. Targeted at research workflows, competitive intelligence, and deep analysis that would take a human analyst days.
Comet — AI Browser (2026)
Perplexity's standalone AI browser. Available on desktop and Android. Comet embeds an AI assistant alongside your browsing — voice chat across tabs, cross-tab summarisation, built-in ad blocking, and live financial data with price heatmaps. Pro subscribers can choose the underlying model (including Opus 4.6 as default). Comet positions Perplexity as a browser company as well as a search product.
Sonar — Perplexity's Own Models
Perplexity built their own model family (Sonar) optimized for real-time grounded answers:
- Sonar — built on Meta Llama 3.3 70B; powers all free-tier queries. Fast, grounded, accurate.
- Sonar Pro — deeper multi-search reasoning for complex queries
- Sonar Reasoning Pro — chain-of-thought reasoning with live web grounding
- R1 1776 — Perplexity's own reasoning model based on DeepSeek R1, fine-tuned for improved factual grounding and uncensored answers
Plans & Pricing
| Plan | Price | Key inclusions |
|---|---|---|
| Free | $0 | Basic search, Sonar model, limited Pro Search queries |
| Pro | $20/mo ($200/yr) | 300+ Pro Searches/day, model choice (GPT-4o, Claude, Gemini), file uploads, image generation, $5/mo API credit |
| Max | $200/mo | Unlimited access to all advanced models, no volume caps. Includes Perplexity Computer — autonomous multi-model agent for long-running tasks. |
| Enterprise Pro | $40/seat/mo | SSO, shared Spaces, admin + audit controls, centralised billing |
| Enterprise Max | $325/seat/mo | Max toolset, higher storage/bandwidth, dedicated support |
Sonar API
The Sonar API gives programmatic access to Perplexity's search-grounded models. Pricing is token-based (input + output tokens). As of 2026, citation tokens are no longer billed on standard Sonar and Sonar Pro — reducing the effective cost per response. Pro subscribers get a $5/month recurring API credit.
When to Use Perplexity
Strong use cases
- • Current events and recent information
- • Research that needs citations you can verify
- • Competitive intelligence and market research
- • Technical questions needing up-to-date docs
- • Any query where training cutoff matters
Not ideal for
- • Long creative writing or document drafting
- • Code generation and debugging
- • Private/internal data questions
- • Multi-turn conversational tasks
Checklist: Do You Understand This?
- How is Perplexity different from asking ChatGPT the same question?
- What is the difference between Pro Search and Deep Research?
- What is the Sonar model family and who builds it?
- What did the Model Council feature (Feb 2026) add?