'Outperforming DeepResearch': New Chinese AI agent Manus rewrites autonomy rulebook. What sets it apart?

Just weeks after DeepSeek grabbed headlines, a new AI agent called Manus is drawing comparisons to some of the most advanced autonomous AI systems from OpenAI, Google, and Anthropic. Developed by Chinese startup Monica, Manus is being pitched as a system that “bridges minds and actions” — an AI that doesn’t just process information but delivers complete results. Launched on March 6, Manus has already captured global attention, with its creators claiming it outperforms OpenAI’s DeepResearch on the GAIA benchmark.

What is Manus?

"Manus is a general AI agent capable of handling complex, real-world tasks autonomously. Unlike traditional AI chatbots, it doesn’t just assist users — it plans, executes, and delivers end-to-end results."

For instance, if asked to write a research paper on global warming, Manus won’t merely generate text. It will research sources, write the paper, code charts and interactive elements, and bundle everything—without needing more user prompts. The same goes for creating a stock analysis, generating an interactive course, or crafting a detailed travel itinerary. Manus claims it can do all this from a single prompt.

A demo video shared by Monica shows the AI working in real time, collecting data and performing tasks live. While OpenAI’s DeepSearch has similar research features, early adopters say Manus is the most advanced autonomous AI system they’ve seen.

Key features  

Autonomous task execution: Manus keeps working in the cloud once it’s given a task, even if the user disconnects. In the demo, it operated across 50 different screens, drawing information from sources like X (formally twitter) and Telegram, and even taking screenshots along the way.  
Real-time interaction & workflow display: Manus shows its progress as it browses the web, uses tools, and compiles results. In one demo, the AI created a day-by-day Japan trip itinerary, revealing each step of its process live.  
Personalization: Like other AI models, Manus adapts to user behavior over time, tailoring its output to individual needs.  
Runs in the background: Even if a user logs off, Manus continues working and sends a notification once the task is done.  
Beyond text generation: Manus not only produces written answers but also browses websites, interacts with live pages, tracks its own activity, and can create downloadable PDFs, spreadsheets, or presentations.

Availability and how to use it

Manus is currently accessible through an invitation-only web preview, with no confirmed public release date. Growing buzz suggests a broader launch is on the horizon, and Monica plans to open-source the model soon, allowing developers to integrate Manus into their projects.

Using Manus resembles platforms like ChatGPT or Grok. You type in a request — say, planning a week-long Bali trip on a tight budget — and Manus autonomously researches, organizes data, and delivers structured, real-time results.