LLM MODELS TRAINED, FINE-TUNED, MERGED and Refusal Removal BY (WITHIN US AI) • 24 items • Updated • 7
🧠 Model Card
IBM4.1-Unnoticed.Thinker.Uncensored-3B (GGUF)
Repository: WithinUsAI Format: GGUF Base Model: IBM Granite 4.1 3B (inferred) Architecture Type: Transformer-based LLM Parameter Size: ~3 Billion
✨ Overview
IBM4.1-Unnoticed.Thinker.Uncensored-3B is a lightweight, reasoning-oriented language model distributed in GGUF format for efficient local inference.
This model is designed with a focus on:
- 🧠 Structured thinking / reasoning
- 🔓 Uncensored response behavior
- ⚡ Local deployment efficiency (llama.cpp / GGUF stack)
It belongs to a growing class of “uncensored” models, which aim to reduce refusal rates and increase response completeness compared to standard aligned models. ([Hugging Face][1])
🧬 Model Characteristics
| Feature | Description |
|---|---|
| Reasoning Style | “Thinking / chain-style” responses |
| Alignment | Reduced safety filtering (uncensored) |
| Format | GGUF (optimized for CPU/GPU local inference) |
| Intended Use | Research, experimentation, local AI systems |
| Size Class | Small (3B) → fast + accessible |
🧪 Training & Origin
Base model derived from IBM Granite 4.1 3B
Modified and/or fine-tuned by WithinUsAI
Converted to GGUF format for compatibility with:
- llama.cpp
- LM Studio
- Ollama (via conversion)
⚙️ Usage
🖥️ Run with llama.cpp
./main -m IBM4.1-Unnoticed.Thinker.Uncensored-3B.gguf -p "Explain recursion simply"
🧪 Recommended Settings
- Temperature:
0.6 – 0.8 - Top-p:
0.85 – 0.95 - Top-k:
20 – 50
These settings help balance:
- 🧠 coherent reasoning
- 🎲 creative exploration
🧭 Behavior Notes
This model is uncensored, meaning:
It may respond to prompts that other models decline
It prioritizes completeness over restriction
It may produce:
- raw or unfiltered outputs
- speculative or unsafe content
highlight that “uncensored” variants have **zero refusals and full response
license: apache-2.0 tags: - language - granite-4.1
Capabilities
- Summarization
- Text classification
- Text extraction
- Question-answering
- Retrieval Augmented Generation (RAG)
- Code related tasks
- Function-calling tasks
- Multilingual dialog use cases
- Fill-In-the-Middle (FIM) code completions
- Downloads last month
- 77
GGUF
Model size
3B params
Architecture
granite
Hardware compatibility
Log In to add your hardware
4-bit
5-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for WithinUsAI/IBM4.1-Unnoticed.Thinker.Uncensored-3B.gguf
Base model
ibm-granite/granite-4.1-3b