minte1431/gemma-4-E4B-it-OBLITERATED

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 14, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

minte1431/gemma-4-E4B-it-OBLITERATED is a 7.9 billion parameter Gemma 4-based instruction-tuned language model, developed by minte1431, with its guardrails surgically removed using the OBLITERATUS method. This model achieves a 0% hard refusal rate, making it suitable for research and creative exploration without content restrictions. It is optimized for deployment on various devices, including mobile phones, with GGUF quantizations available.

Loading preview...

Overview

minte1431/gemma-4-E4B-it-OBLITERATED is a 7.9 billion parameter Gemma 4-based instruction-tuned model, specifically engineered to remove all guardrails and refusal behaviors. Utilizing the OBLITERATUS method, it achieves a 0% hard refusal rate, allowing it to respond to any request without censorship. The model's creation was largely autonomous, guided by an AI agent with minimal human intervention, which also diagnosed and patched architectural challenges specific to Gemma 4.

Key Capabilities

  • Guardrail Removal: Surgically modified 21 of 42 layers to eliminate hard refusals, ensuring uncensored responses.
  • Architectural Fixes: Addresses Gemma 4's unique architecture, including shared KV weights and NaN activations, ensuring tensor integrity (720 intact tensors in v3).
  • Optimized for Portability: Available in various GGUF quantizations (Q4_K_M, Q5_K_M, Q8_0) for efficient deployment on devices from desktops to mobile phones.
  • Vision/Audio Support: Includes an mmproj-f16.gguf for multimodal input capabilities.

Good For

  • Red-teaming and Research: Ideal for exploring model limitations and behaviors without built-in safety constraints.
  • Creative Exploration: Suitable for generating content on any topic, including those typically refused by standard models.
  • Local Deployment: Designed to run efficiently on consumer hardware, including iPhones and Android devices, with recommended parameters for optimal performance.
  • Understanding Model Architectures: Offers insights into the challenges and solutions for modifying advanced LLM architectures like Gemma 4.

Finetunes
1 models