# 4 billion parameter optimization

Ministral is a GPT-like model based on 4 billion parameters, adopting the same architecture as the Mistral model but on a smaller scale, primarily designed for English text generation tasks.