carsenk/llama3.2_1b_2025_uncensored_v2
Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:1BQuant:BF16Ctx Length:32kPublished:Jan 17, 2025License:llama3.2Architecture:Transformer0.0K Warm

The carsenk/llama3.2_1b_2025_uncensored_v2 is a 1 billion parameter Llama 3.2 base model fine-tuned by Carsen Klock with a 32768 token context length. This model is specifically optimized for uncensored responses, medical reasoning, mathematics problem-solving, and code generation. It leverages a diverse dataset including specialized instruction, math, code, and uncensored conversation data. Its primary strength lies in providing direct, unfiltered answers across various technical and sensitive topics.

Loading preview...

Llama 3.2 1B Uncensored Overview

This model, llama3.2_1b_2025_uncensored_v2, is a 1 billion parameter variant of Meta's Llama 3.2, fine-tuned by Carsen Klock. It distinguishes itself through its focus on providing uncensored responses and specialized capabilities in medical reasoning, mathematics problem-solving, and code generation. The model was trained using LoRA fine-tuning with Unsloth, across 79,263 steps.

Key Capabilities

  • Uncensored Responses: Designed to engage with any topic and provide direct, honest answers without moral constraints. Users can activate this mode via a specific system prompt.
  • Specialized Reasoning: Excels in complex domains such as medical reasoning and advanced mathematics, leveraging datasets like medical-o1-reasoning-SFT and math-gpt-4o-200k.
  • Code Generation & Feedback: Proficient in generating code and providing feedback, trained on datasets like CodeFeedback-Filtered-Instruction and dolphin-coder.
  • General Instruction Following: Maintains strong general instruction following capabilities from its diverse training data, including FineTome-100k.

Training Details

The model was fine-tuned on a combination of datasets including FineTome-100k (general instructions), orca-math-word-problems-200k and math-gpt-4o-200k (mathematics), CodeFeedback-Filtered-Instruction and dolphin-coder (code), Jenna-50K-Alpaca-Uncensored (uncensored conversations), and medical-o1-reasoning-SFT (medical reasoning). It supports a 32768 token context length and is available in GGUF (f16, q4_k_m) and merged 16-bit formats.

When to Use This Model

This model is particularly well-suited for applications requiring:

  • Direct and unfiltered responses on a wide range of subjects.
  • Assistance with complex mathematical problems.
  • Code generation, debugging, and feedback.
  • Medical information processing or reasoning tasks.
  • General instruction following where a smaller, specialized model is preferred.
Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p