Mon Agent IA
découvrez comment deepseek, une innovation majeure en intelligence artificielle inspirée du modèle chinois, transforme le paysage technologique. explorez ses fonctionnalités, ses applications et son impact sur l'industrie.

DeepSeek: a significant advance in artificial intelligence based on the Chinese model

Agent Olivier
June 10, 2025

In a constantly evolving technological landscape, artificial intelligence (AI) is taking a predominant place, influencing our daily lives in unprecedented ways. Last week, a major innovation from DeepSeek, a start-up based in Hongzhou, China, captured the world’s attention with the launch of its new model, the DeepSeek R1. As this app quickly makes its way to the top of the app rankings, a shake-up is looming in the tech industry, affecting the current dominance of American AI giants. This phenomenon raises questions about the future of international competition in advanced technologies.

The DeepSeek R1: a revolution in artificial intelligence

Within days of its launch, the DeepSeek R1 saw massive user adoption, quickly becoming the most downloaded app on the Apple App Store. On the Google Play Store, the figures speak for themselves: 10 million downloads in record time. But what makes this app so captivating and why has it had such an impact?

A formidable rival for American models

The DeepSeek R1 isn’t just popular; it rivals and exceeds the performance of state-of-the-art AI models developed by companies such as Meta and OpenAI. Using a battery of tests across 21 standard benchmarks, R1’s results were impressive. In these tests, it managed to outperform its competitors for 12 criteria. This marks a turning point in the perception of China’s technological capacity, which now seems ready to align itself in terms of quality with American leaders.

  • Outperformance on 12 of 21 benchmarks
  • Second place out of 8 additional criteria
  • Best model on several comparative tasks

A technological feat with reduced costs

What is particularly impressive is how DeepSeek reduced the computing resources required to create R1. Comparatively, developing R1 required only 2.8 million computing hours on an infrastructure equipped with NVIDIA H800 graphics cards. In comparison, Meta exceeded 30 million hours for an equivalent model. This savings is linked to an innovative design that could change the way companies approach AI model creation.

Model Computing Hours Required Number of Parameters
DeepSeek R1 2.8 million 671 billion
Meta Llama-3.1 30.8 million 405 billion

Open Source: A New Collaboration Model

DeepSeek has also broken new ground by making R1 available as an open-source model. This decision allows anyone to download, inspect, and modify the model. AI developers and researchers appreciate this flexibility, allowing them to adapt R1 to diverse applications. This approach stands in stark contrast to that of platforms like OpenAI, which keep their models closed.

Implications of the Open Source Model

In a world where collaboration is essential, DeepSeek’s choice to opt for an open source model could revolutionize the AI ​​landscape. Not only does it broaden the horizon of possibilities for developers, but it also encourages competitive dynamics. Previous success stories demonstrate that knowledge sharing can lead to major innovations:

  • Development of diverse solutions leveraging R1 by third parties.
  • Specific adaptations to address local or niche problems.
  • Strengthening collaborative research capabilities on a global scale.

Chain of thought and transparency

Another notable aspect is the implementation of a “chain of thought” procedure. Unlike OpenAI models, which hide the reasoning steps, R1 allows users to visualize each step of the decision-making process. This transparency strengthens user confidence in the results produced.

Global Economic Impact: The US Reaction

The launch of DeepSeek R1 is not just a technological breakthrough; it also has a considerable economic impact. Wall Street reacted vigorously, with an estimated $1 trillion loss in the value of American companies in one day. This shockwave calls into question American strategies for dominance in the AI ​​sector, particularly in a context where massive investments are being announced.

The Rise of the StarGate Initiative

In the days following the DeepSeek announcement, President Trump introduced the StarGate initiative, aimed at investing $500 billion in AI technology in the United States. The goal is to establish a resilience against growing competition from China. With large-scale AI data center projects, the geopolitical landscape of technology could undergo very significant changes.

The Stakes of Technological Dominance

The implications of DeepSeek’s rise go beyond the market. They touch on the very essence of the technological confrontation between the United States and China. With companies like Huawei, Alibaba, and Tencent continually pushing the envelope, the rivalry for leadership in artificial intelligence is more intense than ever. The United States must redouble its efforts to maintain its position.

Company Investment (in billions USD) Focus
DeepSeek Undisclosed Open Source AI, Consumer Application
StarGate LLC 500 AI Data Centers
Open AI Undisclosed Closed AI Models

Future Outlook and Opportunities

As the AI ​​industry evolves, DeepSeek R1 could be the catalyst for new opportunities. Companies across various sectors, such as Meituan, Didi Chuxing, and ByteDance, could leverage this model to enhance their services. It also paves the way for new startups to emerge, borrowing from R1’s features and architecture.

Transforming Business Models

With a cost of use significantly lower than competing models, DeepSeek is also revolutionizing industry dynamics by making AI more accessible. Companies that choose to use R1 will not only be able to reduce their costs but also increase their agility and innovation capacity. Lower cost of use compared to OpenAI (e.g., 27 times cheaper)

  • Access to an open-source model for specific adaptations
  • Potential to develop new services tailored to user needs
  • The importance of adaptability in a changing digital world

For companies like NIO or Xiaomi, integrating AI models like R1 could be crucial to remaining competitive in a rapidly evolving global market. Adaptability is often the key success factor in a dynamic business landscape. Brands that capitalize on these innovations will be well positioned to thrive in the future.

In this environment of continuous innovation, the ability to compete and collaborate will determine which players gain the upper hand in the years to come. Thanks to tools like DeepSeek R1, we are at the dawn of a new era in artificial intelligence.