mlabonne/Llama-3.1-70B-Instruct-lorablated

The mlabonne/Llama-3.1-70B-Instruct-lorablated is a 70 billion parameter instruction-tuned language model based on Meta's Llama 3.1 architecture, developed by mlabonne. This model is specifically designed to be an uncensored version of Llama 3.1 70B Instruct, achieved through a LoRA-abliteration process. It maintains high quality while removing censorship, making it suitable for general-purpose applications and role-play scenarios where unconstrained responses are desired.