Mistral Unleashes 1-Trillion-Parameter Beast

🔥 What happened Mistral released the weights of ML4 on October 29, 2026 – internally dubbed „Le Chonk,“ a 1-trillion-parameter model with 49 billion active per token. Mistral claims it’s the best open-weight model for coding and cybersecurity. 💡 Why it matters ML4 is three times smaller than Kimi K3 and runs in FP4 quantization on just four GB300 GPUs – a budget universities can afford. Mistral trained it in its own French data centers using only ten megawatts, which it calls „ridiculously little“ compared to competitors. ⚡ Our take Mistral is betting big on cybersecurity and open weights – a smart move against US and Chinese models. Whether „Le Chonk“ truly matches Qwen and Gemma will require independent benchmarks.