FiditeNemini/Unhinged-Qwen2.5-R1-1M-Uncensored-BF16

TEXT GENERATIONConcurrency Cost:1Model Size:14.8BQuant:FP8Ctx Length:32kPublished:Jan 28, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

FiditeNemini/Unhinged-Qwen2.5-R1-1M-Uncensored-BF16 is a 14.8 billion parameter causal language model, merged from DeepSeek-R1-Distill-Qwen-14B-abliterated-v2 and Qwen2.5-14B-DeepSeek-R1-1M using the 'ties' method. This model is uncensored and converted to MLX bfloat16 format, featuring a notable 1 million token context length. It is designed for applications requiring extensive context processing and an uncensored response capability.

Loading preview...

Model Overview

FiditeNemini/Unhinged-Qwen2.5-R1-1M-Uncensored-BF16 is a 14.8 billion parameter causal language model, distinguished by its uncensored nature and an impressive 1 million token context length. This model was created through a "ties" merge of two base models: "huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2" and "mkurman/Qwen2.5-14B-DeepSeek-R1-1M".

Key Characteristics

  • Architecture: A merged model combining elements of DeepSeek-R1-Distill and Qwen2.5, specifically designed for extensive context handling.
  • Parameter Count: 14.8 billion parameters, offering a balance between performance and computational requirements.
  • Context Length: Features a significantly extended context window of 1,000,000 tokens, enabling processing of very long inputs and maintaining coherence over extended dialogues or documents.
  • Uncensored: Designed to provide responses without inherent content filtering, offering greater flexibility for specific applications.
  • Precision: Converted to MLX bfloat16 format, optimizing it for MLX-compatible hardware and potentially improving inference speed and memory efficiency.

Use Cases

This model is particularly well-suited for scenarios demanding:

  • Long-form content generation and analysis: Due to its 1M context length, it can handle extensive documents, codebases, or conversations.
  • Applications requiring uncensored outputs: Ideal for research or creative tasks where content filtering might be undesirable.
  • MLX ecosystem integration: Optimized for use within the MLX framework, benefiting from bfloat16 precision for efficient deployment.