[France Mistral releases new generation AI model, targeting the enterprise network security market] October 30th, French artificial intelligence company Mistral AI said on Thursday local time that its latest Large 4 model (ML4 for short) has advanced programming and network defense capabilities, and enterprises can deploy it to run on their own computers and servers. The model was trained in Mistral's data center in Europe, using 3,800 NVIDIA Grace Blackwell GPUs. The company stated that the model has a total parameter volume of 1.05 trillion, and the actual parameters activated each time a task is performed are approximately 52 billion. While ML4 achieves "industry-leading performance," it uses only a "small fraction" of the computing power and capital of closed-source AI companies such as OpenAI and Anthropic.