The new Perplexity computer is another bet that users need many more AI models

Starting this week, Perplexity subscribers will have a new proxy tool at their disposal.

In the company’s words, Perplexity Computer “unifies all existing AI capabilities into a single system.” More specifically, Perplexity says it is a computer user agent that can autonomously execute complex workflows using 19 different AI models, and even create sub-agents to handle specific problems.

The tool is now only available on the company’s highest subscription tier, the $200 per month Perplexity Max. It runs entirely in the cloud, which may avoid some of the security concerns of other proxy tools like OpenClaw.

TechCrunch hasn’t done a hands-on demo of the new tool, but in a workflow example on Perplexity’s website, it’s shown handling tasks that involve collecting statistics, financial or legal data; Create analysis and share its results as websites or ready-made visualizations.

Perplexity invited the press to a press conference with executives last week to discuss the product and set the agenda for the year. The event was intended to include a demo of the tool, but the company canceled the demo due to defects found in the product hours before the event.

The tool represents an evolution of Perplexity, which made a splash early in the AI ​​boom by wrapping pioneering models in familiar user interfaces, particularly its search engine-like answers service. It then moved on to launch its own Comet web browser last summer. Competitors like Google have now changed their products to be more like those built at Perplexity, but that’s as much a threat as it is a compliment, one executive said.

The company is changing in response to a changing ecosystem: It was one of the first AI companies to introduce ads, and it abandoned that business late last year, saying last week that it had undermined users’ confidence in the accuracy of their answers. But Perplexity’s total user base — tens of millions of users — pales in comparison to that of OpenAI, which claims to have 800 million weekly users and began testing ads in ChatGPT this year.

Now, Perplexity executives say they’re aiming for a broader range of users, with products that serve people who make “decisions that move GDP.” Executives at the conference, who requested anonymity, described prioritizing enterprise subscriptions, especially for deep research.

“You don’t hear us talking about monthly active units at all, because we’re not really on a mission to get as many users as possible,” one executive said.

Perplexity recently released a new benchmark for complex research tasks, called Draco, where (not surprisingly) its deep research offerings outperform competitors like Gemini.

Perplexity says it no longer relies on other companies’ APIs for its web index and now has its own AI-optimized search API. But the company is doubling down on its efforts to populate leading models into a consumer-friendly user experience, arguing that there is value in coordinating multiple third-party MAs to get the most accurate and cost-effective answers to queries.

“Multi-models are the future,” argued one Perplexity executive. Models, from their point of view, are specialized, not commodity. The company found that its users frequently switched between templates to get the results they were looking for, with December 2025 queries for visual output most often sent to Gemini Flash, software engineering done in Claude Sonnet 4.5, and medical research done in GPT-5.1.

If one LLM is better at programming tasks and another does a better job crafting marketing copy, Perplexity can automatically select the ideal software. Another example is running modified, open-source LLM software made in China by Perplexity to answer queries at a lower cost, a technology the company was exposed for hiding from its customers last year, executives said. But if this technique is applied transparently, it can be an effective way to improve LLM queries.

The company also provides users with the opportunity to inquire about multiple models at once, in a feature called Model Council. But the unit economics of serving multiple queries at fixed subscription prices are not entirely clear.

However, without expensive infrastructure projects on its books and with high margins on user fees, as executives claimed, Perplexity believes it will remain competitive by allocating tokens to the best model for a purpose.

And there’s more on the horizon: The Perplexity Comet browser is coming to iOS next month, and the company plans to hold a developer conference, Ask, on March 11 in San Francisco to promote third-party use of its API.

One executive said that instead of looking at the number of inquiries from the previous day each morning, he was now looking at the latest revenue metrics. At least some customers are noticing a new focus on the bottom line, with the Perplexity subreddit featuring frequent complaints about new price caps for free and subscription product tiers.

However, executives at the press conference dismissed such complaints, with one saying: “Any discussions about making the free tier worse or restricting the price are completely false.”