Model Overview

huihui-ai/Qwen2.5-14B-Instruct-abliterated-v2 is a 14.8 billion parameter instruction-tuned language model derived from the Qwen/Qwen2.5-14B-Instruct base. Developed by huihui-ai, this version has undergone a process known as "abliteration" to remove inherent censorship, distinguishing it from its base model. This modification aims to provide a more open-ended and less restricted conversational AI experience.

Key Characteristics

• Uncensored Output: The primary differentiator is its abliterated nature, designed to generate responses without the content restrictions typically found in instruction-tuned models.
• Base Model: Built upon the robust Qwen2.5-14B-Instruct architecture, inheriting its multilingual capabilities (supporting languages like Chinese, English, French, Spanish, German, etc.) and instruction-following proficiency.
• Context Length: Features a substantial context window of 131072 tokens, allowing for extended conversations and processing of lengthy inputs.
• Improved Version: This v2 iteration is noted as an improvement over its predecessor, Qwen2.5-14B-Instruct-abliterated.

Use Cases

This model is suitable for developers and applications that require an instruction-following large language model with an emphasis on uncensored content generation. It can be integrated into various applications using the Hugging Face transformers library, with example code provided for conversational use. It is also available for use with Ollama.