Manus vs. DeepSeek: Is Manus China's Next AI Breakthrough?

The Buzz Around Manus: Is the Hype Justified?

The recently launched Manus, an “agentic” AI platform currently in preview, has sparked considerable excitement, rivaling the anticipation surrounding a major event like a Taylor Swift concert.

A product lead at Hugging Face characterized Manus as the most impactful AI tool they have experienced. Furthermore, an AI policy analyst, Dean Ball, described the platform as representing the most advanced AI-driven computing to date.

Rapid Growth and Demand

The official Manus Discord server experienced explosive growth, reaching over 138,000 members within days of its launch.

Access to Manus is highly sought after, with invite codes reportedly being traded for substantial sums – in the thousands of dollars – on Chinese resale platforms like Xianyu.

Underlying Technology and Claims

However, the extent to which this enthusiasm is warranted remains questionable.

Manus isn’t built on entirely original code. Reports suggest the platform leverages a combination of pre-existing and refined AI models, including Anthropic’s Claude and Alibaba’s Qwen, to handle tasks like generating research reports and analyzing financial data.

Despite this, Butterfly Effect – the Chinese startup developing Manus – presents ambitious examples on its website, suggesting the platform can perform complex actions such as real estate purchases and video game programming.

Performance Benchmarks and Comparisons

Yichao “Peak” Ji, a research lead for Manus, indicated in a viral video that the platform surpasses other agentic tools like OpenAI’s deep research and Operator in capabilities.

Ji asserted that Manus achieves superior performance on GAIA, a benchmark used to evaluate the ability of AI assistants to perform tasks involving web browsing and software utilization.

“[Manus] represents a fundamental shift in human-machine interaction,” Ji stated. “It’s not merely another chatbot or workflow; it’s a fully autonomous agent that connects ideas with their implementation.”

Early User Experiences

Initial feedback from some users, however, suggests that Manus doesn’t consistently deliver on its promises.

Alexander Doria, co-founder of AI startup Pleias, reported encountering errors and infinite loops during Manus testing. Other users on X noted inaccuracies in responses and inconsistent citation practices, with the platform frequently missing readily available information.

Personal Testing Results

My own testing of Manus yielded similarly underwhelming results.

I tasked the platform with a seemingly simple request: ordering a fried chicken sandwich from a highly-rated fast food restaurant within my delivery area. After approximately ten minutes, Manus experienced a system crash.

A subsequent attempt identified a suitable menu item, but Manus was unable to complete the order or even provide a link to the checkout page.

Similarly, Manus failed to successfully book a flight from New York City to Japan.

Despite clear instructions specifying preferences for business class, price, and flexible dates, the platform only provided links to various airline websites and airfare search engines, some of which were non-functional.

Further attempts to utilize Manus – including reserving a restaurant table and developing a Naruto-inspired fighting game – resulted in errors and ultimately, a decision to discontinue testing.

Official Response

A spokesperson for Manus provided TechCrunch with the following statement via direct message:

“As a small team, our primary focus is continuous improvement of Manus and the creation of AI agents that genuinely assist users in problem-solving. The current closed beta is designed to rigorously test the system and identify areas for improvement. We greatly value the feedback received from our users.”

Factors Contributing to the Hype

If Manus is currently falling short of its stated capabilities, what explains the widespread attention it has received? Several factors played a role, including the limited availability of invitations.

Chinese media outlets quickly promoted Manus as a significant AI advancement, with QQ News proclaiming it “a source of national pride.” Social media AI influencers also contributed to the hype by disseminating inaccurate information regarding the platform’s functionality.

A widely circulated video depicted a desktop application, falsely presented as Manus, interacting with multiple smartphone apps. Ji later confirmed that this video was not a genuine demonstration of Manus.

Furthermore, influential AI accounts on X drew comparisons between Manus and DeepSeek, a Chinese AI company, despite a lack of factual basis.

Unlike DeepSeek, which develops its models internally, Butterfly Effect relies on existing technologies. Additionally, while DeepSeek has made many of its technologies publicly accessible, Butterfly Effect has not yet done so.

Conclusion

Acknowledging that Manus is still in early access, the company asserts it is actively working to expand computing capacity and address reported issues.

However, based on its current state, Manus appears to be an instance of inflated expectations exceeding actual technological progress.

Updated 6:02 p.m. Pacific: Added a statement from a Manus spokesperson and corrected a misidentification of the company behind Manus.