On January 20, 2025, a Chinese research lab launched a New AI Model called DeepSeek R1. Along with it, they published a research paper claiming that this chatbot outperforms the most advanced AI models in the world, including OpenAI’s ChatGPT o1, Meta’s Llama, and Google’s Gemini Advanced.
The New AI Model R1 is not only the best in terms of performance and efficiency, but it was also developed at a fraction of the cost and time compared to its competitors. While OpenAI charges $200 per month for its premium model. But this is completely free for users. The cost of training this was only $5.6 million, whereas American companies like OpenAI, Meta, and Google are spending billions on their AI models.
Just a week after its launch, This became the most downloaded app in the U.S., surpassing ChatGPT. The next day, it topped the charts in India and other countries. By January 27, it had already shaken up the American stock market. The launch of DeepSeek wiped out $1 trillion from the U.S. tech index. NVIDIA, the world’s most valuable company before DeepSeek’s release, saw its stock price drop drastically, losing $589 billion in a single day—marking the biggest loss in history.
The Story Behind China’s New AI Model
Liang Wenfeng founded DeePseek at 40 and kept a low public profile. He founded a hedge fund called High Flyer in 2015 that based investment decisions on artificial intelligence and mathematics. He founded High Flyer AI in 2019 to work with artificial intelligence algorithms. By May 2023, he had started a side project to build an AI model that was better than anything available, purely out of scientific curiosity rather than profit.
Instead of hiring engineers, he built a team of PhD students from China’s top universities. Over two years, with just a few million dollars, they created DeepSeek R1. Only 200 people worked on it, and 95% of them were under 30 years old. This is in stark contrast to OpenAI, which has over 3,500 employees.
DeepSeek’s Advantages and Limitations
This stands out for its ability to show its reasoning process while answering questions. Unlike previous AI models that provide instant responses, This carefully considers different angles before responding. This makes it highly accurate in logic-based tasks like math and riddles.
Despite this, the open-source nature makes it highly valuable. Anyone can download and modify it, removing censorship if needed. American companies like Perplexity AI and Microsoft have already done this. Microsoft even integrated this R1 into Azure AI Foundry and plans to add it to Copilot.
DeepSeek vs. Other AI Models
When compared to leading AI chatbots:
- DeepSeek excels in coding, quantitative reasoning, and riddles.
- ChatGPT o1 leads in scientific knowledge, poetry, and table creation.
- In terms of response time, ChatGPT is faster (31.15 seconds),
- while DeepSeek takes longer (71.22 seconds), mainly due to high server demand.
- One of DeepSeek’s biggest innovations is its ‘Mixture of Experts’ method.
- Unlike ChatGPT o1, which functions as a single model,
- DeepSeek has separate specialized models for different fields.
For example, if you ask a legal question, it will activate its ‘lawyer’ model instead of a general AI. This makes it more efficient while reducing energy and costs.
The AI Wars Have Begun
After this success, OpenAI accused it of copying proprietary models, claiming they had evidence of ‘distillation,’ a method where smaller AI models learn from larger ones. Ironically, OpenAI itself has been accused of using copyrighted content without permission. Several lawsuits have been filed against OpenAI, including one by The New York Times and 17 famous authors.
Meanwhile, the U.S. government had imposed export restrictions to prevent China from acquiring advanced NVIDIA chips needed for AI training. However, this New AI Model managed to work with older chips and still create a highly efficient model. This innovation has led to concerns that AI development no longer requires billions of dollars and cutting-edge hardware.
Conclusion
DeepSeek’s launch marks a major shift in the AI industry. It proves that world-class AI can be built with fewer resources and at a lower cost. While its censorship is a downside, its open-source nature allows users to modify it. With American companies already integrating it, this is set to challenge the dominance of OpenAI and other tech giants. The AI wars have just begun, and the future looks more competitive than ever.
One Comment on “China’s New AI Model Destroys American ChatGPT”