After two months of relative quiet, Perplexity AI has re-emerged with what it calls its most ambitious product yet - Perplexity Computer. Framed as a 'general-purpose digital worker,' the system is designed to take a high-level objective and execute it end to end, and that includes researching, planning, coding, deploying, monitoring and iterating.
For a company that began as an AI-powered answer engine in 2022, the leap is significant. Perplexity is no longer positioning itself as a smarter search bar of the remaining lot. It is now competing in the fast-evolving rat race to build autonomous AI agents that can manage real-world workflows with minimal supervision.
So, what does it mean for you, or your operations team?
From Answers to Outcomes
As per their blog post, instead of responding to a single prompt, Perplexity Computer breaks a desired outcome into tasks and subtasks, assigns them to specialised sub-agents and executes them asynchronously. Each task runs inside an isolated compute environment equipped with a browser, filesystem access and tool integrations.
If something goes wrong, the system can spawn additional sub-agents to troubleshoot. The coordination layer automatically routes subtasks to the most suitable model from a pool of 19 models.
As of launch, its reasoning engine is powered by Opus 4.6, with integrations including Gemini for research, Nano Banana for image generation, Veo 3.1 for video, Grok for lightweight operations and ChatGPT 5.2 for long-context recall and search.
In effect, users describe an outcome like “build a portfolio dashboard” or “track real estate launches and generate monthly reports”, and the system plans and executes the workflow across models. It is model-agnostic by design, which means it allows users to choose specific models for subtasks and manage token spending through a usage-based credit system.
ALSO READ: AI Tsunami Is Coming: Anthropic's Dario Amodei On Why Society Is Missing The Big Picture
Who Gets Access — And How to Use It
Perplexity Computer is currently available only to Perplexity Max subscribers. Access is web-based, and users can begin by describing a desired outcome in natural language.
Max subscribers receive 10,000 credits per month under a usage-based pricing model. At launch, the company is offering a one-time bonus of 20,000 additional credits for both existing and new Max users. These bonus credits expire 30 days after they are granted.
Users can allocate credits across subtasks, select which models power specific agents and set spending caps to control costs. This flexibility is central to the product's positioning as “massively multi-model.”
Rollout to Pro and Enterprise tiers is expected later, though the company has not provided a firm timeline.
How It Differs From OpenClaw and Claude
Anthropic's Claude models, including the Opus line, are known for strong reasoning and long-context understanding. Claude can write code, analyse documents and reason through complex problems. However, it typically operates within a conversational loop. Even with tool use and memory, it remains largely prompt-driven.
Perplexity Computer shifts the focus from conversation to orchestration. Instead of asking the user for the next instruction at every step, it plans a multi-stage workflow, executes tasks in parallel and manages dependencies across models.
The contrast is sharper with OpenAI's OpenClaw, an open-source autonomous agent that runs locally on a user's machine. OpenClaw offers deep system-level access, connecting to emails, files and applications directly. Developers can configure which models it uses and how much operational control it receives.
Perplexity's approach is more centralised, and cloud-based. Computer runs within the company's managed infrastructure. That means Perplexity controls integrations, imposes safeguards and determines how tasks interact with external systems. For enterprises, this hosted model may offer clearer governance and accountability. For developers who value full autonomy and system-level flexibility, open-source agents like OpenClaw may remain attractive.
Broader Platform Play
Perplexity describes Computer as the next step after Comet, its AI-native browser, and Comet Assistant, its personal agent layer. Persistent memory, task tracking and deep research capabilities are now woven into a single platform.
The company has also signalled that Computer is built for scale. It can manage multiple long-running projects simultaneously and maintain continuity over time. In public demonstrations, the system has been shown researching APIs, writing application code, deploying simple web apps and generating recurring reports without repeated human prompts.
Chief executive Aravind Srinivas described the launch as the result of months of focused development, positioning it as part of Perplexity's broader mission to “power the world's curiosity.”
Perplexity Computer does enter a crowded and fast-moving field. Tech giants and startups alike are experimenting with agents that can plan, act and adapt across digital environments. The challenge is balancing capability with control, and autonomy with safety.
ALSO READ: AI Hidden Load: Can Grids Keep Up With The Model Boom?
Essential Business Intelligence, Continuous LIVE TV, Sharp Market Insights, Practical Personal Finance Advice and Latest Stories — On NDTV Profit.
