Perplexity’s Computer: Betting on Multiple AI Models for Users
▼ Summary
– Perplexity has launched a new agentic tool called Perplexity Computer, which uses 19 AI models to autonomously execute complex workflows and is available only on its $200/month Max subscription tier.
– The company is shifting its strategy to target a boutique, enterprise-focused user base making high-stakes decisions, prioritizing deep research subscriptions over mass user growth.
– Perplexity is emphasizing a multi-model approach, arguing that AI models are specializing and its system can automatically select the best model for specific tasks like coding or medical research.
– The company has abandoned its advertising business, citing a conflict with user trust, and now relies on its own AI-optimized search API instead of other companies’ APIs.
– Despite user complaints about new rate limits, executives deny making the free tier worse and highlight upcoming developments like bringing the Comet browser to iOS and hosting a developer conference.
Subscribers to Perplexity now have access to a powerful new feature called Perplexity Computer, a tool that integrates multiple artificial intelligence capabilities into one streamlined system. This agentic tool autonomously executes complex workflows by leveraging 19 distinct AI models, even generating specialized subagents to tackle specific components of a task. Currently exclusive to the premium Perplexity Max plan costing two hundred dollars monthly, the tool operates fully in the cloud, a design choice the company suggests mitigates certain security concerns associated with similar agentic software.
While a full independent review is pending, demonstrated workflows from Perplexity show the tool gathering statistical, financial, and legal data, performing analysis, and then presenting its conclusions through completed websites or data visualizations. The launch follows a strategic shift for the company, which recently discontinued its advertising business late last year, stating that ads compromised user trust in answer accuracy. Perplexity’s leadership now emphasizes targeting a specialized user base focused on high-stakes, “GDP-moving decisions,” with a clear priority on enterprise subscriptions for deep research.
The introduction of Perplexity Computer marks a significant evolution for the company. Initially gaining attention by integrating advanced AI models into user-friendly, search-like interfaces, Perplexity later launched its Comet web browser. Company executives note that larger competitors have since adjusted their own products to mirror Perplexity’s approach, a trend viewed as both flattering and competitive. The core philosophy driving this new tool is a belief in a multi-model future, where different large language models (LLMs) specialize in distinct tasks rather than becoming generic commodities.
Internal data supports this approach, showing users naturally gravitate toward different models for different needs. For instance, visual generation requests often go to Gemini Flash, software engineering tasks are handled by Claude Sonnet, and medical research queries are directed to GPT-5.1. Perplexity’s software aims to automate this selection, routing queries to the most capable or cost-effective model automatically. This includes utilizing modified open-source models for certain queries to optimize costs, a practice the company faced criticism for last year when it was not fully transparent. Executives now argue that when disclosed, this method is an efficient way to manage resources.
The company also provides a Model Council feature, allowing users to query several models simultaneously for a single request. However, the financial sustainability of offering unlimited multi-model queries under a flat subscription fee remains an open question. Perplexity executives express confidence, citing high margins on user fees and the absence of costly infrastructure projects, allowing them to competitively allocate computational resources.
Looking ahead, Perplexity plans to release its Comet browser for iOS next month and will host a developer conference named Ask in San Francisco to encourage third-party use of its APIs. This commercial focus is becoming more apparent internally; one executive noted a shift from monitoring daily query volumes to tracking revenue metrics. This renewed emphasis on profitability has not gone unnoticed by users, with online forums showing frequent complaints about new rate limits on both free and paid tiers. Company leadership firmly denies these claims, stating any suggestion of degrading the free tier is “completely false.”
(Source: TechCrunch)
