AI manus: Chinese AI capable of doing everything on its own

With our partner Salesforce, unify sales, marketing and customer service. Accele your growth!

The launch of Manus marks a new significant advance in the field of general autonomous agents, coming directly from the Chinese technological ecosystem. Discreetly developed by the company Manus AI, this new artificial intelligence solution is not limited to conventional interactions via a chatbot but acts independently, by itself by piloting all of the tasks assigned to it, without requiring intermediate human intervention.

Manus is distinguished by its ability to fully execute complex projects, from start to finish, through a series of structured steps. Unlike most existing artificial intelligences that require regular human intervention, manus operates independently by first creating a detailed action plan, then by carrying out each task associated with its own internal resources.



Among the demonstrations presented when it was launched, Manus was able to perform various tasks such as:

  • In -depth financial analysis: Manus has demonstrated its ability to produce a complete stock market title. The agent has independently carried out a SWOT analysis, assessed the market position, analyzed its key financial metrics and provided specific investment recommendations.
  • Generation of digital content and programming : Manus is able to automatically create and deploy complete websites. For example, during a public demonstration, he collected data on a public personality, generated a detailed biography, developed an HTML/CSS website corresponding, and deployed it online independently, including the management of possible technical problems.
  • In -depth research and complex project management : Another notable demonstration includes personalized real estate research in New York, where Manus analyzed criminal statistics, districts characteristics, and offered a selection of housing adapted to the precise criteria of the user.

https://www.youtube.com/watch?v=K27DIMBCSU

Technically, manus is based on a multi-agent model using several artificial intelligences specializing simultaneously, orchestrated on a cloud infrastructure operating under Linux. Each task is broken down into sub-objectives treated independently by these different sub-agents, with a contextual memory allowing manus to learn from each interaction. The notable technical particularity is its functioning on an open-source Linux environment, considerably increasing flexibility and speed of execution.

According to the available preliminary assessments, Manus obtains impressive results, including a score of 86.5 % in the Gaia benchmark (level 1), largely surpassing models such as Openai (74.3 %). This performance demonstrates a significant technological advance, placing China in a strategic competitive position in the face of Western leaders in the sector.

Manus remains to date accessible only by invitation, via a waiting list. There is no doubt that his media coverage seems to be deign, his official account on X/Twitter has been disabled.

A few weeks after Deepseek, Manus AI recovered the debates on the competition on the AI ​​chessboard, a match which is far from over.