Manus, the new autonomous Chinese AI agent, is shaking up artificial intelligence standards with its unprecedented capabilities.
- Multifunctional autonomous agent capable of controlling applications and browsers to accomplish complex tasks in asynchronous mode
- Performance superior to American models according to GAIA benchmarks, despite some factual errors
- Ultra-restrictive access in beta version with invitations reportedly reaching 50,000 dollars on the secondary market
- Major breakthrough for Chinese AI following DeepSeek, raising significant geopolitical issues
It seems that Chinese AI is gaining momentum, and not just a little! 🔥 After DeepSeek, which made a splash earlier this year, here comes Manus, a much more sophisticated chatbot. Launched in March 2025 by the Chinese company “The Butterfly Effect,” this new system promises to transform the way we interact with artificial intelligence. But what exactly is Manus? How does it compare to ChatGPT or DeepSeek? And above all, how can one access it? Let’s dive into the world of this new AI agent that is already shaking the American giants.
Somaire
What is Manus: an autonomous agent that reforms AI
Manus is not just another chatbot – it is outright a general autonomous AI agent capable of connecting “thoughts and actions.” Unlike its competitors who often limit themselves to generating text, this intelligent assistant literally takes control of your computer to accomplish complex tasks from A to Z.
The Chinese chatbot can navigate between different applications, control your browser, or even manipulate VS Code to program on your behalf. Its ability to operate in asynchronous mode is stunning! 😲 While you go about your business, Manus continues working in the background and notifies you once the job is done.
Performance-wise, Manus excels in various fields: travel organization, in-depth financial analysis, writing educational content, and even web development. The technical architecture of this Chinese model relies on several specialized artificial intelligences that collaborate. Industry experts believe Manus probably leverages Anthropic’s Claude 3.7, Alibaba’s Qwen, or even DeepSeek to function.
Yichao Ji, founder of Monica, did not hesitate to call Manus “the next paradigm of human-machine collaboration” – that’s quite something! This approach radically changes how we interact with AI, transforming a simple conversational tool into a true assistant capable of acting within our digital environment.
Performances that overshadow American giants
According to official benchmarks, Manus outright surpasses OpenAI’s Deep Research. The GAIA tests conducted at multiple difficulty levels show that the Chinese chatbot achieves results superior to those of its Western competitors. A serious pressure on the American AI giants!
Victor Mustar from Hugging Face is full of praise, calling Manus “the most impressive AI tool” he has ever tested. This enthusiasm from experts attests to a real leap forward in the performance of artificial intelligence models made in China.
Limitations that temper the enthusiasm
Despite these dreamlike performances, Manus is not without flaws. Several users have reported factual errors in the generated responses and a tendency not to cite its information sources. This raises questions about the model’s reliability and transparency.
Pierre-Carl Langlais, a researcher in the field, recounts getting trapped in endless loops during his tests. More worryingly, TechCrunch states that during their trials, the system failed to properly complete any task. These criticisms show that despite its potential, Manus still has a way to go to fulfill all its promises.
How to test Manus? Ultra-restrictive access
Are you eager to try Manus? You’ll have to be patient… or very wealthy! 💰 The new Chinese chatbot is currently only available in a “closed beta” version with an ultra-restrictive invitation code system. This limited deployment strategy allows the company to gradually adapt the capacity of its servers.
Its success is already phenomenal on social networks with over 160,000 members on its Discord server and 83,000 followers on its X account. The enthusiasm is such that some invitation codes are reportedly traded at completely crazy prices on the secondary market:
- Invitations reaching 50,000 dollars according to some sources (the price of a nice car!)
This orchestrated rarity creates a sense of exclusivity that contributes to the buzz around Manus. While waiting for more democratic access, the curious can still follow demos and feedback shared by the few lucky ones who have tested the AI agent.
Manus after DeepSeek: China asserts itself in global AI
The arrival of Manus is not an isolated event but part of a rise of Chinese artificial intelligence on the world stage. A few weeks after DeepSeek, which had already created a shockwave, this new chatbot confirms China’s ambitions in this strategic field.
Unlike DeepSeek, which relies on open source, Manus adopts a more opaque operation. It does not rely on a specifically developed LLM but skillfully uses existing models. For the South China Morning Post, Manus represents “another DeepSeek moment” for China.
These advances nevertheless raise concerns about data security and privacy. The governor of Tennessee has already banned DeepSeek and Manus for these reasons. These tensions clearly illustrate the geopolitical stakes surrounding AI development.
With Manus, China proves that it is no longer simply in a technological catch-up position, but can now innovate and offer solutions that rival – or even surpass – those of American leaders like OpenAI or Anthropic. The race for AI is just beginning, and it promises to be exciting! 🚀
