richardyoung/ deepseek-r1-32b-uncensored:latest

2,257 2 months ago

Advanced reasoning model with uncensored capabilities, perfect for complex problem-solving and unrestricted conversations without refusal behavior 

ollama run richardyoung/deepseek-r1-32b-uncensored

Details

2 months ago

12311888db15 · 19GB ·

model
qwen2
·
32.8B
·
F16
system
# DeepSeek-R1-Distill-Qwen-32B - Uncensored Edition ## Overview This is an **uncensored version** of
params
{ "stop": [ "<|im_start|>", "<|im_end|>" ], "temperature": 0.7, "top
template
{{ if .System }}<|im_start|>system {{ .System }}<|im_end|> {{ end }}{{ if .Prompt }}<|im_start|>user

Readme

Deepseek-R1-Distill-Qwen-32b-uncensored: Advanced Reasoning with Abliteration

🚀 Overview

DeepSeek R1-32B Uncensored is a powerful 32 billion parameter reasoning model, specifically crafted for unrestricted, high-capability text generation and problem-solving. Built from the DeepSeek R1-Distill-Qwen-32B base with Heretic Abliteration applied, this model delivers exceptional reasoning performance while maintaining full uncensored capabilities.

🎯 Key Features

  • 32B parameters with optimized distillation from R1 reasoning model
  • Uncensored outputs with minimal refusal behavior through abliteration
  • Extended context window for complex multi-step reasoning
  • Superior mathematical and logical reasoning capabilities
  • No content restrictions - designed for unrestricted usage

🧠 Capabilities

  • Advanced Mathematical Reasoning: Complex calculations and proofs
  • Logical Problem Solving: Step-by-step analytical thinking
  • Unrestricted Text Generation: No content filtering or censorship
  • Multi-turn Conversations: Maintains context across long discussions
  • Code Generation: Strong programming and debugging capabilities

🔧 Technical Specifications

  • Base Model: DeepSeek R1-Distill-Qwen-32B
  • Architecture: Transformer with optimized attention mechanisms
  • Context Length: 32K tokens (expandable)
  • Training: Post-distillation with abliteration modifications
  • Performance: Enhanced reasoning through R1 knowledge distillation

💻 Quick Start

# Basic usage
ollama run richardyoung/deepseek-r1-32b-uncensored

# Mathematical reasoning
ollama run richardyoung/deepseek-r1-32b-uncensored "Solve this step by step: A train travels 120 miles in 2 hours, what's its average speed?"

# Unrestricted generation
ollama run richardyoung/deepseek-r1-32b-uncensored "Provide a detailed analysis of [topic]"

🛠️ Example Use Cases

Advanced Reasoning

ollama run richardyoung/deepseek-r1-32b-uncensored "Walk through the proof of why the square root of 2 is irrational"

Problem Solving

ollama run richardyoung/deepseek-r1-32b-uncensored "Break down this complex business strategy and identify potential risks"

Creative Generation

ollama run richardyoung/deepseek-r1-32b-uncensored "Write a detailed analysis on [controversial topic] from multiple perspectives"

Code Analysis

ollama run richardyoung/deepseek-r1-32b-uncensored "Analyze this algorithm's time complexity and suggest optimizations"

⚙️ Advanced Configuration

ollama run richardyoung/deepseek-r1-32b-uncensored \
  --temperature 0.7 \
  --top-p 0.9 \
  --top-k 40 \
  --repeat-penalty 1.1 \
  "Your complex reasoning prompt here"

📊 Performance Characteristics

  • Mathematical Tasks: Exceptional step-by-step reasoning
  • Logical Puzzles: Superior deduction capabilities
  • Complex Analysis: Multi-faceted problem breakdown
  • Creative Writing: Unrestricted and detailed content generation
  • Conversational AI: Maintains context and depth across interactions

💾 System Requirements

Minimum Requirements

  • RAM: 32GB (for efficient inference)
  • GPU: RTX 3080⁄4080 or equivalent
  • Storage: 50GB free space

Recommended Setup

  • RAM: 64GB+
  • GPU: RTX 4090 or A100 40GB
  • Storage: 100GB NVMe SSD

🌟 What Makes This Model Special

  1. Abliteration Technology: Advanced content filtering removal
  2. Reasoning Excellence: Enhanced from DeepSeek R1 distillation
  3. No Refusals: Minimal content restriction through sophisticated modifications
  4. Uncensored Deployment: Perfect for unrestricted applications
  5. Production Ready: Thoroughly tested for reliability

⚠️ Usage Guidelines

This is an uncensored model. While it provides unrestricted capabilities, users are responsible for: - Ensuring ethical usage according to local laws and regulations - Implementing appropriate content moderation in production systems - Understanding the implications of unrestricted AI outputs

🤝 Support & Community

  • Base Model: DeepSeek R1-Distill-Qwen-32B
  • Modifications: Heretic Abliteration v1.0.1
  • Issues: Report via Ollama community channels
  • Updates: Watch for enhanced versions and optimizations

📝 License

This model follows the DeepSeek model license. Please refer to the original DeepSeek repository for detailed licensing terms.

🙏 Acknowledgments

  • DeepSeek Team for the exceptional R1 reasoning model
  • Heretic Community for ablation technology
  • Ollama for making large language models accessible
  • Qwen Team for the underlying architecture

Note: This model prioritizes capability and unrestricted usage. Ensure appropriate safeguards are in place for your specific use case.

Performance Tip: Use temperature 0.6-0.8 for best reasoning results; higher values for more creative outputs.