Welcome Le Chonk: Mistral Large 4 on Hugging Face

Mistral Large 4 is a multimodal, multilingual, Mixture-of-Experts (MoE) model with approximately 1 trillion total parameters and a 1 million token context window. It is particularly strong in cybersecurity and coding, accepting both text and images as input to generate text. The model's architecture features a Pixtral vision encoder and a text decoder that uses sparse attention mechanisms and a combination of 128 experts. Released on Hugging Face under an open-weight license, it supports various inference libraries like Transformers and vLLM and can be fine-tuned using TRL.