Mistral Unveils Open-Source Plan, Aims to Outperform DeepSeek with New AI Model

March 9, 2025 – French company Mistral AI, often referred to as the “European OpenAI,” has announced plans to embrace an open-source strategy, aiming to release a new AI model that surpasses DeepSeek. This announcement was made by the company’s CEO, Arthur Mensch, during the 2025 World Mobile Communication Conference.

According to Mensch, the adoption of an open-source approach will enable the company to develop more powerful AI technologies at a lower cost. He believes that this strategy will not only facilitate rapid iteration and improvement in the company’s AI offerings but also promote collaboration within the global AI ecosystem.

“We are planning to release a highly advanced model in the near future, one that will outperform DeepSeek,” Mensch stated. He views DeepSeek as a contributor to the open-source community rather than a competitor, emphasizing that the open-source model fosters healthy competition and mutual learning among enterprises, ultimately driving the overall progress of AI technology.

Meanwhile, in the United States, a joint AI venture named “Stargate” was announced in January, aiming to establish a data center for OpenAI. The initial investors in this project include OpenAI, SoftBank Group, Oracle, and MGX, while Arm, Microsoft, and NVIDIA serve as technical partners. With an initial investment of 100billionandplanstoinvestupto500 billion in the next four years, Stargate represents a significant commitment to AI development.

However, Mensch highlighted that the open-source model adopted by Mistral AI allows the company to leverage resources effectively, eliminating the need for massive investments to develop cutting-edge AI models. “We don’t believe it takes $100 billion to train an excellent model,” he emphasized. This approach underscores the efficiency and cost-effectiveness of open-source development in advancing AI technology.

Leave a Reply