nguyenthilaitrieulong/phi-4-abliterated

TEXT GENERATIONConcurrent Unit Cost:1Model Size:14.7BQuant:FP8Context Size:32kPublished:Aug 4, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

The nguyenthilaitrieulong/phi-4-abliterated model is a 14.7 billion parameter language model based on the microsoft/phi-4 architecture, fine-tuned to remove refusal behaviors. This model utilizes an abliteration technique, a proof-of-concept method for uncensoring LLMs without TransformerLens. With a context length of 32768 tokens, it is designed for use cases requiring less restrictive content generation.

Loading preview...

Model Overview

This model, nguyenthilaitrieulong/phi-4-abliterated, is a 14.7 billion parameter language model derived from the microsoft/phi-4 architecture. Its primary distinguishing feature is the application of an "abliteration" technique, a proof-of-concept method aimed at removing refusal behaviors and uncensoring the base LLM without relying on TransformerLens.

Key Capabilities

  • Uncensored Content Generation: The model is specifically modified to reduce or eliminate refusal responses, allowing for broader content generation.
  • Base Architecture: Built upon the microsoft/phi-4 foundation, inheriting its general language understanding and generation capabilities.
  • Context Length: Supports a substantial context window of 32768 tokens, enabling processing of longer inputs and generating more coherent extended outputs.

Use Cases

  • Exploratory Research: Ideal for researchers investigating methods of model censorship removal and behavioral modification.
  • Unrestricted Text Generation: Suitable for applications where the default refusal mechanisms of standard LLMs are undesirable or counterproductive.
  • Ollama Integration: Directly compatible with Ollama (version 0.5.5 or newer) for easy local deployment and experimentation.