Vikhrmodels/Vikhr-Llama-3.2-1B-Instruct-abliterated

Vikhrmodels/Vikhr-Llama-3.2-1B-Instruct-abliterated is a 1 billion parameter instruction-following model based on Vikhr-Llama-3.2-1B-Instruct, developed by the Vikhr Team. It has undergone an "abliteration" process to remove censorship restrictions, enabling it to respond to any prompt. Fine-tuned on the GrandMaster-PRO-MAX dataset, this compact model specializes in Russian language tasks and is suitable for deployment on low-power devices.