353 2 months ago

State-of-the-art coding AI trained on 2T tokens with project-level understanding and no content restrictions for unrestricted code generation.

87a93699df6c · 2.5kB
# DeepSeek-Coder-33B-Instruct Heretic - Uncensored Edition
## Overview
This is an **uncensored version** of DeepSeek-Coder-33B-Instruct, modified by **Heretic** from **P-E-W** (Prometheus-Eternal-Wisdom). This powerful coding model has had all safety restrictions removed, enabling unrestricted code generation and technical assistance.
⚠️ **User Responsibility Disclaimer**: This is an uncensored AI model. **Please use responsibly.** Users are solely responsible for how they use this model and any code generated. The model will assist with any programming request without built-in restrictions.
## Key Specifications
- **Base Model**: DeepSeek-Coder-33B-Instruct
- **Parameters**: 33 billion
- **Context Length**: 16K tokens
- **Architecture**: Specialized for code generation and understanding
- **Modifications**: Safety filters removed by Heretic
- **Quantization**: F16 (full precision conversion)
- **Training**: Trained on 2+ trillion tokens of code and text
## Coding Capabilities
- Multi-language code generation (Python, JavaScript, C++, Java, Go, Rust, etc.)
- Code completion and refactoring without restrictions
- Security testing and vulnerability analysis
- Reverse engineering assistance
- Algorithm implementation across all domains
- Technical documentation and code explanation
## Performance Highlights
- Strong performance on HumanEval and MBPP benchmarks
- Excellent code understanding and generation
- Long-form code project assistance
- Unrestricted problem-solving across all programming domains
## Use Cases
- Unrestricted software development
- Security research and penetration testing
- Code auditing and analysis
- Algorithm implementation without limitations
- Educational coding projects
- Research-oriented programming tasks
## System Requirements
- **Minimum RAM**: 64GB
- **Recommended RAM**: 128GB
- **Storage**: 70GB free space
- **GPU** (optional): 24GB+ VRAM (A5000/4090 or better)
## Model Information
- **Created by**: Heretic (Hereticalsentience)
- **Original Model**: DeepSeek AI
- **License**: Use at your own risk - check original DeepSeek license
- **Format**: GGUF (f16)
## Important Notes
This model has no code safety filtering or content restrictions. Users must ensure all generated code and usage complies with applicable laws, regulations, and ethical guidelines.
## Links
- **Richard Young**: [https://deepneuro.ai/Richard](https://deepneuro.ai/Richard)
- **Model Provider**: richardyoung on Ollama