Model Overview

Nabbers1999/gemma-3-27b-it-abliterated-refined-novis is an experimental 27 billion parameter instruction-tuned model derived from Gemma 3. It originates from a biprojected norm-preserving abliteration of the base Gemma 3 27B Instruct model, specifically Nabbers1999/Gemma-3-27B-it-NP-Abliterated.

Key Characteristics

• Response Neutrality: This variant has undergone additional fine-tuning specifically to encourage neutral responses. Early testing indicates no significant behavioral differences between toxic and harmless prompts, even without a system prompt.
• Text-Only: The model's vision features have been explicitly removed, reverting it to a purely text-based language model.
• Experimental Nature: It is important to note that this is an experimental model and does not incorporate any inherent safety features. Users are responsible for its deployment and outputs.

Intended Use Cases

This model is suitable for developers and researchers interested in:

• Exploring models with a focus on response neutrality and consistent behavior across diverse inputs.
• Applications requiring a text-only LLM where vision capabilities are not needed.
• Experimental setups where the absence of built-in safety features is understood and managed by the user.