The race to develop advanced AI agents capable of working autonomously has intensified with a new contender from China. Manus AI has emerged on the global stage, claiming to have created the world's first truly autonomous AI agent that can perform complex tasks with minimal human supervision.
The Emergence of Manus AI
Manus AI, developed by Chinese startup Butterfly Effect, has launched an early preview version of what it describes as the first general AI agent. According to Yichao Peak Ji, chief scientist for Manus AI, this technology represents a significant leap beyond conventional chatbots or workflow tools. In an introductory video, Ji emphasized that Manus AI delivers results with minimal human prompting, positioning it as a bridge between conception and execution. The name Manus derives from the Latin motto Mens et Manus, meaning Mind and Hand, symbolizing the connection between thought and action.
Capabilities and Performance Claims
The agent's capabilities extend beyond simple interactions to include complex real-world tasks. According to its website, Manus can create custom travel plans, research real estate properties for affordability, and perform correlation analyses between stocks. In a demonstration, Ji showcased the agent screening 15 resumes, ranking candidates, evaluating each one, and organizing the information into a spreadsheet. A key feature highlighted is the agent's asynchronous cloud-based operation, allowing users to close their laptops while Manus completes assigned tasks. The company claims that Manus has achieved state-of-the-art performance on the GAIA Benchmark, which evaluates general AI assistants, putting it alongside OpenAI's Deep Research agent.
Claimed Capabilities:
- Resume screening and candidate evaluation
- Custom travel planning
- Real estate property research
- Stock correlation analysis
- Performing tasks on freelance platforms like Upwork and Fiverr
Technical Foundation and Development
While specific technical details remain limited, Ji has revealed that Manus operates as a multi-agent system powered by different AI models. The agent builds upon the open-source community's work, with Ji acknowledging that its capabilities wouldn't be possible without these foundations. Recent statements from Ji indicate that Manus's offerings are based on Claude and fine-tuned versions of Alibaba Group's Qwen models. This approach likely allows the company to develop advanced capabilities at a fraction of the cost required to build cutting-edge systems from scratch. The team plans to open-source some of their models later this year, potentially providing more insight into their technical approach.
Key Features of Manus AI:
- Operates as a multi-agent system
- Works asynchronously in the cloud
- Can receive new instructions while working
- Offers standard and high-effort processing modes
- Based on Claude and fine-tuned versions of Alibaba's Qwen models
Market Reception and Expert Opinions
The launch of Manus AI has generated significant interest, with invitation codes reportedly being resold for thousands of dollars on China's reseller app Xianyu. Expert reactions have been mixed, with some AI researchers expressing profound admiration while others remain skeptical. Victor Mustar, head of product at Hugging Face, called Manus the most impressive AI tool I've ever tried, praising its mind-blowing agentic capabilities. Dean Ball, an AI research fellow at George Mason University, suggested that unlike DeepSeek's earlier breakthrough, which replicated capabilities already achieved by American firms, Manus is actually advancing the frontier.
Limitations and Criticisms
Despite the excitement, early user experiences have revealed several limitations. Some users report that the service is slow and occasionally crashes before completing tasks, likely due to limited computing resources. Professor Derya Unutmaz, who compared Manus with OpenAI's Deep Research, found that while Deep Research completed his task in under 15 minutes, Manus failed after 50 minutes without finishing all necessary steps. Alexander Doria, co-founder of French AI lab PlelAs, dismissed the hype, describing Manus as fundamentally a workflow rather than an actual agent beyond the built-in agentic capacities of existing models like Claude. Yiran Chen, an electrical and computing engineering professor from Duke University, characterized Manus as a half-finished product that may be rushing to market to attract investors.
Global AI Competition Implications
Manus AI's emergence follows DeepSeek as the second Chinese AI breakthrough to capture global attention in recent months. Both developments are raising questions about the perceived U.S. lead in artificial intelligence development. With Butterfly Effect reportedly having raised over $10 million in financing from prominent VC firms like ZhenFund and HongShan, the company appears positioned to compete globally. CEO Xiao Hong has explicitly stated ambitions to compete internationally, not just within China. Unlike some U.S. counterparts and DeepSeek, Manus has not published detailed papers about its technology or released code for independent evaluation, making it difficult to fully assess its innovations.
Future Prospects and Industry Impact
As AI agents continue to evolve, Manus represents an important milestone in the development of autonomous systems capable of performing complex tasks with minimal human oversight. Ji has framed Manus as the next paradigm of human-machine collaboration, and potentially a glimpse into AGI (artificial general intelligence). The company is currently advertising 20 job openings with annual salaries reaching as high as CNY 560,000 (approximately USD 77,130) for engineers, suggesting plans for significant expansion. Whether Manus can overcome its current limitations and fulfill its ambitious claims remains to be seen, but its emergence signals an intensifying global race to develop increasingly autonomous AI agents that can transform how humans interact with technology.