.png){kind=logo}
The AI landscape observes a ferocious rivalry between two Chinese giants, DeepSeek and Manus AI. Both throw stones, but the strategies they use are different worlds apart. Hangzhou-based startup DeepSeek left the world with affordable but high-performance models like R1 that threatened to overpower Western behemoths.
Monica AI group's Manus AI launches a new rival with agentic power that plans to overcome its compatriot. This struggle defines the path of AI dominance and questions what technology will reign.
In January 2025, DeepSeek unveiled R1, a reasoning model that competes with OpenAI's o1. It reduces computational costs by only activating a small portion of its 670 billion parameters per task, which is built on a mix-of-experts architecture. Reports peg its training at under $6 million, a fraction of the $100 million-plus spent on GPT-4, leveraging Nvidia H800 chips despite U.S. export curbs.
Open-sourcing the code of R1 guarantees that its potential remains as large as possible and offers developers worldwide to tailor and apply it with unlimited freedom. DeepSeek shatters because it's affordable and widely available, smashing the misconception of AI as something hungry for resources.
Manus AI enters the battlefield with a clear intention: doing, not thinking. Introduced in March 2025 as an invite-only beta, this utility app, created by Butterfly Effect, carries out complex tasks such as writing websites, buying and selling shares, and organizing trips. Unlike DeepSeek's word-focused ability, Manus crosses idea and execution, asserting superiority on the GAIA benchmark for general AI assistants.
Its cloud-based platform brings together models such as Anthropic's Claude and Alibaba's Qwen, optimized for real-world applications. This agential move indicates the trend toward AI that not only thinks but acts, disrupting DeepSeek's analytical dominance.
DeepSeek's strength is efficient engineering. Its multi-head latent attention minimizes memory consumption, and reinforcement learning avoids human feedback, lowering costs. Performance is high in math, programming, and reasoning, equalling or surpassing Western models. Manus, on the other hand, values flexibility.
Its architecture supports real-time task execution, adapting to changing inputs like finances or trip plans. Less informative about its base, demonstrations have it displaying fluency for automation of workflow. DeepSeek balances for understanding; Manus prioritizes action—two schools of thought competing to become the AI style of the future.
DeepSeek's launch shook markets, wiping out $600 billion from Nvidia's value in a day as investors wondered if they needed expensive chips. Its app was the number one download in the U.S., indicating widespread acceptance. Manus, though newer, evokes the same fervour.
Invite codes are worth thousands on Chinese platforms, and its Discord hit 138,000 members in days. Stock surges in AI-concept firms followed its release, echoing DeepSeek's earlier ripple. Both entities break stories of Western dominance, but Manus's practical bent may sway industries looking for tangible results over theoretical innovations.
DeepSeek is challenged on its cost, $6 million in training, not total R&D, and potential data privacy issues with Chinese servers. Bans in Australia and Italy are a manifestation of geopolitical tensions. Manus, however, is plagued by instability. Early adopters experience crashes while performing activities such as booking flights, implying hype outstrips refinement.
Its use of existing models raises questions about originality, and scalability has yet to be proven. Both need to carve through technical fineness and international trust to guide, as DeepSeek's effectiveness conflicts with Manus's ambition.
DeepSeek and Manus AI are changing the face of AI, both with their own flavour of brilliance. Put DeepSeek on the mind as that smart, frugal buddy who has discovered the key to powerful reasoning without sacrificing costs: yes, actually, who doesn't love a deal?
And then there's Manus, the aggressive pioneer, leaping directly into the driver's seat with an autonomous implementation that's straight out of a science fiction movie. The question on everyone's mind is: who will be the winner? It comes down to execution—DeepSeek's ability to scale up against Manus's pursuit of smoothing out the wrinkles and producing bulletproof reliability.
