Manus, an AI Agent product launched by Chinese startup Monica, has recently attracted widespread attention on the X platform. Its founder Ji Yichao introduced the technical background of the product in detail through social media. It is reported that Manus is developed based on Alibaba's Qwen big model and has achieved its unique functions through a variety of fine-tuning models. This disclosure not only sparked industry discussions about the source of Manus technology, but also further stimulated strong interest from the global AI community in its potential.
Ji Yichao posted on the X platform: "In the process of building Manus, we adopted Claude3.5Sonnet v1 and different versions of Qianwen fine-tuning model. Initially, we only had Claude3.5Sonnet v1 available, so we needed to rely on a large number of auxiliary models. Now, Claude3.7 performs well, we are conducting internal testing, and will update progress soon." This statement clarifies that Manus is not a completely independently developed model, but is optimized and integrated based on existing big models, which is in sharp contrast to the previous speculation that some people regard it as a "new breakthrough in China's AI."
As an AI Agent product, Manus' core function is to independently perform complex tasks. According to its official website, the system can not only generate text, but also complete the entire process from planning to execution, such as screening resumes, analyzing stock trends and even generating websites. X user kimmonismus said after the trial: "Manus' preview version far exceeds OpenAI's Deep Research in terms of tasks, which is shocking!" Another user ai_for_success shared a demonstration video showing Manus' smooth operation of browsing web pages in real time, collecting data and outputting results, saying it "looks very cool and can't wait to experience it."
However, with the enthusiastic response from users, doubts also followed. Some users pointed out that Manus relies on existing models rather than self-developed technologies, which may limit its innovation. After testing, X user Pleias founder Alexander Doria said he had encountered problems with error messages and infinite loops, suggesting that its stability still needs to be improved. In addition, due to the current invitation-based access, the insufficient server capacity has caused many users to be unable to experience it, further aggravating the outside world's speculation about its real capabilities.
The Monica team positioned Manus as a "universal AI agent connecting thinking and action", trying to achieve breakthroughs through multi-model collaboration and tool integration. Ji Yichao once emphasized in the demonstration video: "This is not an ordinary chat robot, but an agent that can complete tasks independently." Despite this, its comparison with self-developed models such as DeepSeek still caused controversy. DeepSeek challenges Western AI giants with low cost and high performance, while Manus is more like an integrated product with completely different technical paths.
At present, Manus' invitation code has been hyped to a high price on second-hand trading platforms. There are reports that the price has soared from 999 yuan to 50,000 yuan, reflecting the market's strong expectations for it. Industry analysts believe that regardless of the originality of technology, Manus' rapid popularity has proved that the influence of Chinese AI startups on the global stage is increasing. As testing advances and possible open source plans, whether Manus can deliver on its promises may be revealed in the coming months.