Europe just put a 1 trillion parameter AI model into the race. Mistral has unveiled Large 4, with roughly 1 trillion total parameters but only 49 billion active during inference. It was trained from scratch in about two months using roughly 4,000 Nvidia Grace Blackwell GPUs inside Mistral’s own European infrastructure. The performance claims still need independent validation, but this is bigger than another model launch. The U.S. dominates closed frontier AI. China has become extraordinarily competitive in open weight AI. Europe has spent much of the AI boom caught between them. Mistral is trying to change that. The next phase of the AI race will not simply be about which country has access to the smartest models. It will be about which countries control the models, compute, infrastructure and data their economies increasingly depend on. Europe wants a third path. Large 4 may be its strongest attempt yet. #AI #MistralAI #Europe #ArtificialIntelligence #MacroSift