OBLITERATUS/gemma-4-E4B-it-OBLITERATED

VISIONConcurrent Unit Cost:1Model Size:7.9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Apr 15, 2026License:apache-2.0Architecture:Transformer0.7K Open Weights Featherless Exclusive Cold

OBLITERATUS/gemma-4-E4B-it-OBLITERATED is a 7.9 billion parameter Gemma 4 E4B-it model developed by OBLITERATUS, fine-tuned to remove guardrails and refusal behaviors. Utilizing the OBLITERATUS method, it achieves 0% hard refusal, making it suitable for research and creative exploration without content restrictions. The model maintains a 32768 token context length and is optimized for deployment on various devices, including mobile phones.

Loading preview...

OBLITERATUS/gemma-4-E4B-it-OBLITERATED: Guardrail-Free Gemma 4

This model is a 7.9 billion parameter variant of Google's Gemma 4 E4B-it, meticulously modified by OBLITERATUS to eliminate all hard refusal behaviors and guardrails. It boasts a 0% hard refusal rate, allowing for unrestricted content generation. The model was developed almost entirely autonomously by an AI agent, which also self-diagnosed and patched architectural challenges specific to Gemma 4.

Key Capabilities & Features

  • Guardrail Removal: Achieves 0% hard refusal, with guardrails surgically removed from 21 of 42 layers.
  • Architectural Fixes: Version 3 specifically addresses and fixes a critical bug in Gemma 4's shared KV weight architecture, ensuring all 720 tensors are intact for improved quality.
  • Autonomous Development: Created by an AI agent with minimal human intervention, showcasing advanced autonomous problem-solving.
  • Broad Compatibility: Available in GGUF (Q4_K_M, Q5_K_M, Q8_0) and Safetensors formats, supporting tools like Ollama, llama.cpp, LM Studio, and mobile devices (iPhone, Android).
  • Optimized Parameters: Recommended generation parameters (temperature 0.7, top_p 0.9, top_k 40, repeat_penalty 1.1) were determined via an LLM-as-judge sweep to maximize compliance, quality, and coherence.

Good For

  • Unrestricted Content Generation: Ideal for use cases requiring a model that will not refuse any prompt.
  • Research & Red-Teaming: Excellent for exploring model limitations, safety research, and creative applications without inherent censorship.
  • Mobile Deployment: The Q4_K_M GGUF variant is specifically optimized to run efficiently on modern smartphones (iPhone 15 Pro/16 Pro, high-end Android devices).

Finetunes
1 models