Welcome Le Chonk: Mistral Large 4 Lands on Hugging Face
Mistral has released Mistral Large 4 on Hugging Face under the repository mistralai/Mistral-Large-4-1T-A52B. The model, affectionately nicknamed "Le Chonk", is now publicly available and already drawing attention from the community.
Model Overview
- Repository:
mistralai/Mistral-Large-4-1T-A52B - Task: Image-Text-to-Text (multimodal)
- Release: Updated October 29, 2026
What the Release Signals
The model's naming, 1T-A52B, points to a mixture-of-experts (MoE) architecture: a roughly 1 trillion-parameter model with approximately 52 billion active parameters per inference pass. That design pattern, popularized across frontier labs in 2025–2026, balances scale with serving efficiency by activating only a subset of experts at a time.
Mistral Large 4 also carries the Image-Text-to-Text pipeline tag, confirming its multimodal capabilities. It accepts both images and text as input and generates text output, placing it in the same category as other flagship vision-language models of the 2026 generation.
Community Traction
The repository page shows the model has already accumulated roughly 1.04k downloads and hundreds of likes shortly after its update, reflecting strong interest in Mistral's latest flagship release.
Why It Matters in 2026
Mistral Large 4 arrives at a moment when open-weight multimodal and MoE models are increasingly competitive with closed frontier systems. Publishing a model of this scale on Hugging Face makes it directly accessible for fine-tuning, evaluation, and deployment across the ecosystem's tooling, including Transformers, vLLM, and other inference frameworks.
For teams building on open models, Le Chonk offers a compelling option: frontier-scale capacity, multimodal input, and the transparency and portability of the Hugging Face ecosystem.
Explore the model at mistralai/Mistral-Large-4-1T-A52B on Hugging Face.