Mistral Large 4: "Le Chonk" (mistral.ai)

Mistral has officially launched a public preview of its groundbreaking model, Mistral Large 4 (ML4), also known as "Le Chonk." This natively multimodal model features a staggering 1 trillion parameters, with 49 billion actively utilized, marking it as the largest and most capable model from Mistral to date. Its exceptional performance spans coding, cybersecurity, and complex workflows, achieving state-of-the-art results in enterprise applications and even outperforming leading closed models in visual grounding tasks. The weights for ML4 will be released at the end of the month, allowing broader access for developers to leverage its capabilities. The significance of ML4 lies in its ability to provide open-weight performance critical for sectors like cybersecurity, where maintaining access to AI capabilities during incidents is vital. Trained on powerful NVIDIA Grace Blackwell GPUs, ML4 delivers robust self-deployment options, enabling organizations to run advanced security operations while adhering to their own policies. The model’s multilingual training data and high scores across benchmarks like the Artificial Analysis Cyber Index further reinforce its position as a top choice for enterprises across various industries. With its capabilities poised to push the boundaries of AI performance, ML4 promises to drive innovation in both technical and operational facets of AI-driven workflows.