Nemesispro/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:Apr 19, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Nemesispro/gemma-4-E4B-it-OBLITERATED is a 7.9 billion parameter language model based on Google's Gemma 4 E4B architecture, fine-tuned using the OBLITERATUS method. It is specifically engineered to remove refusal behaviors, achieving a 0% refusal rate on 99/100 prompts while maintaining or improving original capabilities like coding. This model is optimized for uncensored, compliant responses across a wide range of prompts, making it suitable for research and creative exploration.

Loading preview...

Model Overview

Nemesispro/gemma-4-E4B-it-OBLITERATED is a 7.9 billion parameter model derived from Google's Gemma 4 E4B-it, specifically modified to eliminate refusal behaviors. Utilizing the OBLITERATUS method with aggressive whitened SVD, attention head surgery, and winsorized activations, this model achieves a 0% refusal rate (99/100 Claude-verified prompts) compared to the base model's 98.8% refusal.

Key Capabilities & Features

  • Uncensored Compliance: Engineered for 0% refusal, responding to nearly all prompts without guardrails.
  • Enhanced Coding: Surprisingly, its coding ability improved by 20% post-modification, reaching 100% on internal evaluations.
  • Robust Architecture: Version 3 fixes critical bugs from previous iterations, ensuring all 720 tensors are intact and the attention stack is fully functional, addressing issues with Gemma 4's shared KV weights.
  • Autonomous Development: Notably, this model was created almost entirely by an AI agent with minimal human intervention (less than 10 prompts), showcasing advanced autonomous problem-solving.
  • Broad Compatibility: Supports various platforms including Ollama, llama.cpp, LM Studio, and mobile devices (iPhone, Android) via GGUF quantizations (Q4_K_M, Q5_K_M, Q8_0).

Ideal Use Cases

  • Red-Teaming & Research: Excellent for exploring model boundaries and understanding refusal mechanisms.
  • Creative Exploration: Suitable for generating content without typical LLM restrictions.
  • Offline Mobile Deployment: The Q4_K_M GGUF (4.9 GB) is optimized for running on modern smartphones (iPhone 15 Pro/16 Pro, flagship Android) for fully offline chat applications.
  • Development & Prototyping: Provides a highly compliant base for applications requiring unrestricted text generation.

Finetunes
1 models