An AI answer engine that started as a cited search alternative and has evolved into an agentic work platform. Its core Search product returns conversational answers with numbered citations, deep research, and access to premium sources (PitchBook, Wiley). Perplexity Computer (Feb 2026) is an autonomous agent system that decomposes tasks into subtasks, spawns sub-agents, operates a real browser and file system, and can run workflows for hours or months—orchestrating multiple frontier models (Opus 4.6 for reasoning, Gemini for research, Veo for video, Grok for fast tasks). Comet is its AI-native browser. Portable Computer runs locally on Windows (NVIDIA RTX 24GB+ VRAM) and Mac without consuming credits. SPACE, its agent sandbox, is open source. Pricing: Pro/Max subscriptions with credits for agentic work; Enterprise Pro $40/seat/mo, Enterprise Max $325/seat/mo (API credits separate). Annualized revenue exceeded $750M in 2026; Nvidia reportedly in talks to invest at $30B+ valuation. APIs: Orchestrator, Agent, Search, Embeddings, Decisions.

Pros & Cons:

Pros: The citation-first design remains the strongest differentiator—every answer links to numbered sources with trust labels, which is still rare among AI search tools; Computer’s multi-model orchestration is genuinely ambitious, routing tasks to specialized models (Opus for reasoning, Gemini for research, Veo for video) rather than forcing one model to do everything; Portable Computer’s local execution (no credits consumed, data stays on device) addresses the privacy concern that plagues cloud-only agents; the API suite is unusually complete—Orchestrator (OpenAI-compatible), Agent (with tools), Search, Embeddings, Decisions—covering both simple queries and production integrations; SPACE being open source adds credibility to its agent security claims and lets developers reuse the sandbox.

Cons: The product surface has exploded—Search, Computer, Comet, Portable, APIs, Enterprise—and understanding which component to use for which task requires significant onboarding; credits-based agent pricing is unpredictable—Computer workflows can consume credits quickly, and the shift from subscription to usage-based billing drew attention as a revenue driver; Enterprise and API billing are separate, so organizations need to manage two cost centers, and Enterprise seats do not include API access; Portable Computer has hardware requirements (NVIDIA RTX 24GB+ VRAM on Windows), excluding many users; the company’s rapid valuation growth and Nvidia investment talks introduce strategic uncertainty—priorities may shift as it scales toward a potential IPO.

Who is it for?:Researchers, analysts, and knowledge workers who need cited answers and deep research with premium sources; professionals who want an agent that can execute multi-step workflows (document generation, data processing, API calls) rather than just answer questions; developers who need a search/agent API with transparent pricing and OpenAI compatibility; privacy-conscious users who want local agent execution via Portable Computer (with compatible hardware); enterprises that need per-seat agent access with admin controls. Not suited for users seeking a simple search box (the product is now much broader), budget-sensitive individuals who only need occasional search (free tier exists but agent features consume credits), or teams wanting a single unified subscription that includes API access.