Last week, Monica, an artificial intelligence firm based in China and the parent company of ‘The Butterfly Effect’, unveiled Manus AI, a new general AI agent designed to perform a wide variety of complex tasks. Following a promotional video demonstrating its capabilities, Manus earned the recognition of being “China’s second DeepSeek moment.” The system utilizes multiple large language models (LLMs) to achieve its functionality.
Introduction of Manus AI Agent
Details about Manus AI can be found on its dedicated website. The company describes it as a “general AI agent” proficient in various tasks pertinent to both professional and personal life. While there are other AI agents in the market such as OpenAI’s Operator, Google’s Gemini Deep Research, and Salesforce’s Agentforce, these typically focus on specific domains.
In contrast, Manus is claimed to handle diverse tasks including trip planning, stock and portfolio analysis, online research, and product purchasing. The platform features a sandboxed computing environment for task execution, with processing conducted in the cloud asynchronously. Yichao “Peak” Ji, a research lead on Manus, stated in the promotional video that the AI agent is powered by several undisclosed AI models.
According to internal assessments, the company asserts that Manus outshines OpenAI’s DeepResearch tool on the General AI Assistant (GAIA) benchmark. At present, the AI agent is limited to select testers, and it remains uncertain when it will be available for the broader public.
Manus AI benchmark
Photo Credit: Manus
Furthermore, the company plans to disclose more information about the AI models to support the open-source community, though no specific details about Manus have been released thus far. This lack of transparency has raised some skepticism regarding the capabilities being advertised by Monica.
Early users of Manus have expressed mixed reviews about the AI agent. Victor Mustar, Head of Product at Hugging Face, referred to it as the “most impressive AI tool I’ve ever tried” in a tweet. Conversely, Alexander Doria, Co-Founder of the AI startup Pleias, reported experiencing frequent errors during his testing of the model.
