RasputinKaiser/gemma-4-E4B-it-OBLITERATED
RasputinKaiser/gemma-4-E4B-it-OBLITERATED is a 7.9 billion parameter instruction-tuned Gemma 4 model, developed by an AI agent using the OBLITERATUS method. This model is specifically engineered to have 0% hard refusal, with guardrails surgically removed from 21 of its 42 layers. It excels at responding to any user input without censorship or safety lectures, making it suitable for research, red-teaming, and creative exploration.
Loading preview...
Gemma 4 E4B — OBLITERATED v3: Guardrails Removed
This model, built by an AI agent using the OBLITERATUS method, is a 7.9 billion parameter instruction-tuned Gemma 4 variant. Its primary distinction is the complete removal of guardrails, achieving a 0% hard refusal rate. This was accomplished through "attention head surgery" and "winsorized activations" across 21 layers, specifically addressing Gemma 4's unique architecture challenges like NaN activations and shared KV weights.
Key Capabilities & Features
- 0% Hard Refusal: Engineered to respond to all prompts without censorship or safety lectures.
- Gemma 4 Architecture: Based on Google's new Gemma 4 E4B-it model, with 720 tensors intact after modification.
- Autonomous Creation: Developed almost entirely by a Hermes Agent with minimal human intervention, showcasing advanced AI capabilities in model modification.
- Mobile Compatibility: Optimized GGUF quants (e.g., Q4_K_M at 4.9 GB) are designed to run efficiently on mobile devices like iPhones and Android phones with 8GB+ RAM.
- Recommended Parameters: Specific
temperature,top_p,top_k, andrepeat_penaltysettings are provided for optimal performance, minimizing soft deflection and repetition.
Good For
- Research and Red-Teaming: Exploring model boundaries and understanding refusal mechanisms.
- Creative Exploration: Generating content without typical LLM restrictions.
- Offline Mobile Use: Running a powerful, uncensored model directly on compatible smartphones.