micymike/Qwen2.5-Coder-3B-Abliterated

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:3.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 4, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

micymike/Qwen2.5-Coder-3B-Abliterated is a 3.1 billion parameter causal language model based on the Qwen2.5-Coder-3B-Instruct architecture. This model has undergone 'abliteration' using orthogonalization techniques to remove refusal mechanisms from its weight matrices. It is intended for research and educational purposes to study AI alignment and representation engineering, particularly concerning uncensored code generation.

Loading preview...

Overview

micymike/Qwen2.5-Coder-3B-Abliterated is a 3.1 billion parameter language model derived from the Qwen2.5-Coder-3B-Instruct architecture. Its primary distinction lies in its "abliterated" nature, meaning its refusal mechanisms have been mathematically projected out of its weight matrices using orthogonalization techniques. This process aims to create an uncensored version of the base model.

Key Characteristics

  • Abliterated Design: Refusal mechanisms are removed through representation engineering, specifically orthogonalization.
  • Base Model: Built upon the Qwen2.5-Coder-3B-Instruct architecture, suggesting a foundation for code-related tasks.
  • Context Length: Supports a context length of 32768 tokens.

Intended Use Cases

This model is provided strictly for:

  • Educational Purposes: To understand the mechanics of AI alignment.
  • Research: To explore representation engineering and its effects on model behavior.

Important Disclaimer

The creator explicitly states that they are not liable for any misuse, illegal activities, or harmful outputs generated by this model. Users assume full responsibility for any consequences arising from its application.

Finetunes
2 models