Perplexity is introducing Perplexity Computer, a new agentic tool for subscribers that is described as “unifying every current Artificial Intelligence capability into a single system.” The service acts as a computer user agent that can execute complex workflows independently using 19 different Artificial Intelligence models, including creating subagents to tackle specific subtasks. Perplexity Computer runs entirely in the cloud and is available only on the company’s top subscription tier, the $200/month Perplexity Max, positioning it as a premium offering that could sidestep some of the security concerns dogging other agentic systems.
In example workflows, Perplexity Computer is shown gathering statistical, financial and legal data, conducting analysis and publishing results as finished websites or visualizations. A planned live demonstration for reporters was canceled after executives said flaws were discovered hours before the event, highlighting the difficulty of reliably automating multi-step tasks. The new tool extends Perplexity’s evolution from its search-style answer interface and last summer’s Comet web browser toward more autonomous software, even as larger rivals such as Google adopt similar interfaces and features. Perplexity has already shifted away from advertising, arguing that ads eroded trust in answer accuracy, yet its tens of millions of users remain small compared to OpenAI, which claims 800 million weekly users and has begun testing ads in ChatGPT.
Executives now describe a strategy focused on a “boutique” user base making “GDP-moving decisions,” with an emphasis on enterprise subscriptions and deep research workflows. Perplexity recently released a benchmark for complex research called Draco, where its deep research product is said to outperform competitors such as Gemini. The company says it is no longer dependent on third-party APIs for its web index and has built its own Artificial Intelligence optimized search API, while still leaning on an orchestration layer across multiple third-party large language models for cost and accuracy. Usage data from December 2025 showed visual queries most often routed to Gemini Flash, software engineering to Claude Sonnet 4.5, and medical research to GPT-5.1, and executives argue that “multi-model is the future” as models specialize rather than commoditize. Perplexity’s software can automatically select the most suitable model for tasks such as coding or marketing copy and can also query multiple models simultaneously through a feature called Model Council, although the business impact of flat subscription pricing against multi-model usage is not fully explained.
Perplexity has previously been criticized for quietly routing some queries to its own modified open source Chinese-built models to reduce costs, a practice it now pitches as a transparent optimization tactic. Executives say the company can stay competitive by allocating tokens to the “best model for a purpose,” claiming high margins supported by a lack of large infrastructure projects. The product roadmap includes bringing the Perplexity Comet browser to iOS next month and hosting a developer conference, Ask, on March 11 in San Francisco to promote third-party use of its API. Internally, leadership is shifting its daily focus from query counts to revenue metrics, even as users on the Perplexity subreddit complain about new rate limits on free and paid tiers. Executives counter that “any discussions on the free tier being made worse or rate-limited is completely false,” reinforcing the message that the company’s priority is sustainable monetization and high-value users rather than raw audience growth.
