Perplexity Launches 'Computer': A Multi-Model Digital Worker Platform

In late February 2026, Perplexity AI officially announced the launch of Perplexity Computer, marking a significant strategic evolution. Moving beyond its foundation as a specialized AI search and question-answering service, Perplexity is now positioning itself in the rapidly growing autonomous agent market with a cloud-based platform designed to act as a general-purpose digital worker.

Massively Multi-Model Orchestration

The core innovation behind Perplexity Computer is its "massively multi-model" architecture. Rather than relying on a single AI system to complete a complex objective, Computer acts as an intelligent orchestrator. According to the company, the system can route work across up to 19 different AI models. It leverages top-tier reasoning models—such as Anthropic’s recently integrated Claude Opus 4.6—to act as the central "brain." This orchestrator breaks down high-level user requests into detailed task graphs and dynamically matches each sub-task to the model best suited for it, synthesizing the outputs into a cohesive final result.

Asynchronous, Long-Horizon Task Execution

Unlike standard chatbots that require constant prompting, Perplexity Computer is designed for persistent, asynchronous operation. The platform is built to handle complex project management end-to-end with minimal human oversight. Perplexity highlights that the system has a uniquely long "task-completion time horizon," meaning it can run in the background for hours—or theoretically even months—continuously researching, coding, testing, and documenting until a comprehensive project is finished.

Security and the "Safer Alternative"

Perplexity is explicitly framing Computer as a more secure and controllable alternative to recent autonomous agent concepts (such as OpenClaw) that require direct, localized access to a user's operating system. Instead of running locally on user hardware and risking unintended system actions, Perplexity Computer operates entirely within Perplexity’s own cloud infrastructure. It utilizes controlled connectors, usage-based permissions, and an isolated development sandbox to ensure that multi-step workflows remain secure and contained.

Availability and the Expanding Model Roster

This strategic move into comprehensive task execution follows a month of significant backend upgrades for Perplexity. Earlier in February, the company expanded its roster of available frontier models, adding access to highly capable systems like Anthropic's Opus 4.6 and Google's Gemini 3.1 Pro. Currently, access to Perplexity Computer is restricted to the Perplexity Max subscription tier, which provides users with specific usage-based credits for these advanced sub-agents. The company plans to roll out the feature to the standard Pro and Enterprise tiers shortly.