Perplexity subscribers now have access to a new agentic tool called Perplexity Computer. The company describes it as a system that unifies every current AI capability. Specifically, it is a computer user agent that can execute complex workflows independently using nineteen different AI models, even creating subagents to handle specific problems.
The tool is available now, but only on the highest subscription tier, Perplexity Max, which costs two hundred dollars per month. It runs entirely in the cloud, a design that might spare it some of the security concerns associated with other agentic tools.
While there has been no hands-on demo by outside press, example workflows on Perplexity’s website show the tool handling tasks like collecting statistics, financial data, or legal information, creating analyses, and sharing its findings as finished websites or visualizations. The company invited press to a background briefing last week to discuss the product, but cancelled a planned demonstration due to flaws discovered in the tool hours before the event.
This launch represents an evolution for Perplexity. The company first gained attention by wrapping advanced AI models in familiar, search-engine-like interfaces. It later launched its Comet web browser last summer. Competitors like Google have since changed their products to be more like Perplexity’s, which executives see as both a threat and a compliment.
The company is adapting to a shifting ecosystem. It was one of the first AI companies to offer advertising but abandoned that business late last year, stating it undermined users’ trust in answer accuracy. Perplexity’s total user base is in the tens of millions, which pales in comparison to OpenAI’s claimed eight hundred million weekly users.
Now, Perplexity executives say they are targeting a more boutique set of users, with products for people making significant economic decisions. They described prioritizing enterprise subscriptions, particularly for deep research, and stated they are not focused on maximizing user counts.
The company recently released a new benchmark for complex research tasks called Draco, where its own deep research offering outperforms competitors. Perplexity also says it is no longer reliant on other companies’ APIs for its web index and now has its own AI-optimized search API. The company is doubling down on packaging frontier models in a consumer-friendly experience, arguing there is value in orchestrating multiple third-party large language models to get the most cost-effective and accurate answers.
Executives argued that multi-model systems are the future, believing models are specializing rather than becoming commodities. The company has observed its users frequently switch between models for different tasks, such as sending visual queries to Gemini Flash, using Claude Sonnet for software engineering, and utilizing GPT for medical research.
Perplexity’s software can automatically choose the ideal model for a given task. Another technique involves running modified open-source models to answer queries more cheaply, a method the company was criticized for hiding from customers last year. Done transparently, executives believe this can efficiently optimize queries.
The company also offers a feature called Model Council, which allows users to query multiple models at once. The economics of offering multiple queries under flat subscription rates are not entirely clear. However, executives claim high margins on user fees and believe the company will remain competitive by allocating resources to the best model for each purpose.
More developments are planned. The Perplexity Comet browser is coming to iOS next month, and the company is planning a developers’ conference called Ask on March eleventh in San Francisco to promote third-party use of its API.
An executive noted a shift in focus from daily query counts to revenue metrics. This new focus on the bottom line has been noticed by some customers, with frequent complaints on the Perplexity subreddit about new rate limits on both free and subscription tiers. Executives dismiss these complaints, stating that any discussion of the free tier being made worse is completely false.