After DeepSeek shockwave, China strikes again with ManusAI: Should ChatGPT, Google Gemini be scared?

China is raising the stakes with consecutive AI releases, the latest being an AI agent from Monica, a Shenzhen-based AI startup. Launched on March 6, 2025, the autonomous AI agent has quickly gained immense popularity, with many calling it the return of another "DeepSeek moment." Touted by the company as the first "general-purpose AI agent," ManusAI, the name by which it is called, is demonstrating capabilities that can challenge the likes of OpenAI-led ChatGPT and Google Gemini.

What is Manus?

Manus, derived from the Latin word for "hand," is a general AI agent designed to transform thoughts into actions. The company describes Manus as an AI agent that serves as a bridge between ideas and execution. "It doesn’t just think; it delivers results. Manus excels at various tasks in work and life, completing everything while you rest."

Last week, the company launched an early preview of Manus, calling it the first general AI agent and claiming it is a "truly autonomous agent" that eliminates the gap between conception and execution.

What can ManusAI do?

ManusAI has rapidly gained popularity due to its key features, including full autonomy, multi-domain capabilities, high-precision personalization, and real-time operation. Unlike traditional AI models that require specific instructions for specific tasks, ManusAI claims to proactively execute tasks and analyze data without human intervention.

"Now, let me show you Manus in action. I've just sent Manus a zip file containing 10 résumé documents. Since each Manus session has its own computer, it operates like a human: first unzipping the file, then browsing through each résumé page by page, and recording important information into documents. Manus works asynchronously in the cloud, meaning you can close your laptop at any time, and Manus will notify you when everything is complete. You can also give Manus new instructions at any time," says Yichao "Peak" Ji, Co-founder and Chief Scientist of ManusAI.

The AI agent can also conduct deep research. For example, a demonstration video shows Manus filtering New York properties based on multiple criteria by first breaking them down and creating a to-do list. It then searches for and carefully reads articles about the safest neighborhoods, writes a Python program to calculate the budget, and finally compiles all the gathered information into a detailed report.

ManusAI can also perform complex tasks, such as correlation analysis between stocks, with precision and can access authoritative data sources via APIs.

Where does ManusAI stand?

On GAIA, a benchmark for evaluating general AI assistants in solving real-world problems, the company claims Manus has achieved new state-of-the-art (SOTA) performance across all three difficulty levels. Evaluated using the same configuration as its production version for reproducibility, Manus reportedly outperforms OpenAI and previous leading models, with accuracy rates of 86.5%, 70.1%, and 57.7% at Level 1, Level 2, and Level 3, respectively.

"It’s only getting better," says Peak. "Beyond benchmarks, Manus has been solving real-world problems on platforms like Upwork and Fiverr and has demonstrated its capabilities in Kaggle competitions."

Why has ManusAI gained popularity?

Though currently available on an invite-only basis, early users are already calling it the arrival of a true "AI agent."

"Manus, the new AI product everyone’s talking about, is worth the hype. This is the AI agent we were promised—Deep Research + Operator + Computer Use + Lovable + Memory. Asked it to ‘do a professional analysis of Tesla stock,’ and it completed what would normally take two weeks in just an hour!" said Deedy Das, Principal at Menlo Ventures.

Rowan Cheung, the founder of The Rundown AI, shared his experience: "For my first test, I asked Manus to create a biography on Rowan Cheung and deploy a website based on that biography. It was insanely impressive to watch it go through my social channels, browse articles, and deploy the site. And it was 100% accurate, with information up to date as of today."

Victor Mustar, Head of Product at Hugging Face, also praised ManusAI, calling it "the most impressive AI tool" he has ever used. "The agentic capabilities are mind-blowing, redefining what's possible. The UX is what so many others promised... but this time, it just works."

Criticism & China’s rise in global AI

Despite its early success, some critics argue that ManusAI is merely a wrapper model running on top of Dario Amodei-led Anthropic’s Claude Sonnet model. Detractors claim the AI agent offers no real advantage and has gained initial popularity solely due to the agentic capabilities of another foundational AI model. Some users also reported that the AI failed to complete tasks consistently.

"Running OpenAI Deep Research and ManusAI side by side! Climbing toward peak intelligence and experiencing cognitive overload! ...Deep Research finished in under 15 minutes. Unfortunately, ManusAI failed after 50 minutes at step 18/20! It was performing quite well, I was watching Manus’ output, and it seemed excellent. However, rerunning the same prompt is frustrating as it takes too long!"

Other criticisms include issues with web search capabilities, the invite-only access model, and limited user availability.

This is not the first time a globally competitive AI agent has emerged from China. Soon after DeepSeek shook up the U.S. AI landscape in January, e-commerce giant Alibaba unveiled its newest AI model, Qwen 2.5 Max, claiming it outperformed DeepSeek’s R1 model and even the likes of OpenAI’s GPT-4o and Meta’s Llama.

DeepSeek’s AI model is said to rival advanced AI models from American companies at a fraction of the cost. Unlike OpenAI and Google’s AI models, which require around $100 million to train, DeepSeek’s model reportedly cost just $5.6 million. VC and adviser to former U.S. President Donald Trump, Marc Andreessen, called the development "AI’s Sputnik moment" and "one of the most amazing and impressive breakthroughs I’ve ever seen."