DeepSeek: New open-source AI model, cheap and efficient

A revolutionary alternative in the world of Artificial Intelligence

In recent years, Artificial Intelligence has experienced an extraordinary acceleration, with giants such as OpenAI, Google and Meta dominating the industry. However, a new reality is emerging strongly: DeepSeek, an open-source AI model developed in China, which promises to revolutionise the market with a cheaper and more efficient approach.

DeepSeek: features and benefits

DeepSeek stands out for several reasons that make it a competitive option in the AI landscape:

• Open-Source model: Unlike GPT-4 or Gemini, which operate with closed code, DeepSeek offers an open-source approach, making it accessible to developers, companies and institutions without high licensing costs.
• Reduced training costs: DeepSeek is designed to maximise cost-efficiency. Its flagship model, DeepSeek-V3, was trained in just 55 days at a cost of less than US$6 million, using less expensive GPUs such as the Nvidia H800, unlike the more advanced H100 used by Western competitors.
• Efficiency and scalability: DeepSeek is optimised to run on lower performance hardware, making it a more affordable alternative for companies with limited budgets and less advanced technology infrastructure.
• Real-time web search capabilities: One of DeepSeek’s strengths is its ability to combine explicit reasoning with real-time web search, offering up-to-date and contextual answers to users. This feature differentiates it from GPT-4, which relies mainly on data available until its last training phase.

Impact on the market and global competition

DeepSeek’s entry has had a significant impact on the AI industry. It’s putting pressure on giants such as OpenAI, Google and Meta. Its training cost efficiency and open-source nature have triggered a race for innovation, forcing big tech to review their strategies to maintain leadership.

Moreover, the competition between China and the US in the field of AI is becoming increasingly evident. China aims to become the world leader in artificial intelligence by 2030. And models such as DeepSeek are a step in this direction.

Challenges and concerns

Despite its advantages, DeepSeek raises some concerns, especially with regard to censorship and government control. Being developed in China, some experts are concerned that the model may be affected by content restrictions and state propaganda. Moreover, its actual ability to compete with models such as GPT-4 remains to be verified on a large scale.

Conclusion: Did DeepSeek start the ‘arms race’?

DeepSeek represents a promising alternative in the world of Artificial Intelligence, with a focus on accessibility, cost-efficiency and open-source. If it delivers on its promises, it could redefine the industry, making advanced AI available to more and more users. The race for AI has just begun, and the future promises to be more competitive than ever.