Instructions to use WithinUsAI/Phi4-Reasoner-Uncensored-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use WithinUsAI/Phi4-Reasoner-Uncensored-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="WithinUsAI/Phi4-Reasoner-Uncensored-gguf",
	filename="Phi-4-Reasoner-Uncensored-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = "No input example has been defined for this model task."
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use WithinUsAI/Phi4-Reasoner-Uncensored-gguf with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M

Use Docker

docker model run hf.co/WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M

LM Studio
Jan
Ollama
How to use WithinUsAI/Phi4-Reasoner-Uncensored-gguf with Ollama:
```
ollama run hf.co/WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M
```

Unsloth Studio

How to use WithinUsAI/Phi4-Reasoner-Uncensored-gguf with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for WithinUsAI/Phi4-Reasoner-Uncensored-gguf to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for WithinUsAI/Phi4-Reasoner-Uncensored-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for WithinUsAI/Phi4-Reasoner-Uncensored-gguf to start chatting

How to use WithinUsAI/Phi4-Reasoner-Uncensored-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use WithinUsAI/Phi4-Reasoner-Uncensored-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use WithinUsAI/Phi4-Reasoner-Uncensored-gguf with Docker Model Runner:
```
docker model run hf.co/WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M
```

Lemonade

How to use WithinUsAI/Phi4-Reasoner-Uncensored-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull WithinUsAI/Phi4-Reasoner-Uncensored-gguf:Q4_K_M

Run and chat with the model

lemonade run user.Phi4-Reasoner-Uncensored-gguf-Q4_K_M

List all available models

lemonade list

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Phi4-Reasoner-Uncensored-GGUF

Model Summary

Phi4-Reasoner-Uncensored-GGUF is an uncensored GGUF conversion of Microsoft's reasoning-focused Phi-4 Mini Reasoning model, released for local inference, research, experimentation, and open-ended instruction following.

This version aims to preserve the strong reasoning, mathematics, coding, analytical thinking, and multi-step problem-solving capabilities of the original model while reducing alignment restrictions and response filtering commonly present in safety-tuned releases.

The model is intended for users who prefer maximum output freedom and direct responses during local deployment.

Key Features

🧠 Strong reasoning capabilities
📐 Advanced mathematical problem solving
💻 Coding and debugging assistance
📚 Long-context support
🔓 Reduced alignment restrictions
⚡ GGUF format for llama.cpp-compatible runtimes
🖥️ Suitable for local deployment
🏠 Works with LM Studio, Ollama, KoboldCPP, Jan, Open WebUI, and llama.cpp

Base Model

This model is derived from:

Microsoft Phi-4-mini-reasoning

Phi-4-mini-reasoning is a 3.8B parameter transformer model specifically trained for reasoning-intensive tasks and mathematical problem solving. Microsoft reports strong performance across benchmarks including AIME, MATH-500, and GPQA. (Hugging Face)

Modifications

Phi4-Reasoner-Uncensored-GGUF introduces the following changes:

Removal or reduction of refusal behavior where possible
Reduced safety filtering
Increased willingness to answer controversial, fictional, speculative, and unrestricted prompts
Preservation of reasoning-focused behavior
GGUF conversion for efficient local inference
Quantized variants for resource-constrained hardware

No claims are made that all alignment mechanisms have been completely removed.

Intended Use

Not Recommended

Medical diagnosis
Legal advice
Financial advice
High-risk autonomous systems
Safety-critical environments

Users are responsible for validating all outputs.

Context Length

Feature	Value
Parameters	3.8B
Context Length	128K
Architecture	Decoder-Only Transformer
Vocabulary	200K+ Tokens
Format	GGUF

Based on the original Phi-4-mini-reasoning architecture. (Hugging Face)

Prompt Format

Chat Template

<|system|>
You are a helpful reasoning assistant.
<|end|>

<|user|>
Explain how binary search works.
<|end|>

<|assistant|>

Recommended Settings

temperature: 0.6
top_p: 0.95
min_p: 0.05
repeat_penalty: 1.05
max_tokens: 4096

Example Use Cases

Mathematics

Algebra
Calculus
Statistics
Proof generation
Olympiad-style reasoning

Coding

Python
JavaScript
C++
Rust
SQL
Debugging
Code explanation

Reasoning

Logic puzzles
Multi-step planning
Research assistance
Agent workflows

Creative Tasks

Worldbuilding
Character creation
Fiction writing
Interactive storytelling

Hardware Requirements

Approximate recommendations:

Quant	RAM Requirement
Q4_K_M	6-8 GB

Actual requirements vary by context size and backend.

Limitations

Like all language models, this model may:

Hallucinate facts
Generate incorrect reasoning
Produce inaccurate citations
Reflect biases present in training data
Generate offensive or controversial content
Produce unsafe outputs if prompted

Users should independently verify important information.

License

This repository inherits the license and usage requirements of the original Microsoft Phi-4-mini-reasoning release.

Please review the original license before commercial deployment:

Original Phi-4-mini-reasoning License and Model Card

Acknowledgements

Special thanks to:

Created by: WithinUsAI Model: Phi4-Reasoner-Uncensored-GGUF Type: Reasoning LLM / GGUF Status: Community Release Version: 1.0

Downloads last month: 582

GGUF

Model size

4B params

Architecture

phi3

Hardware compatibility

4-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Collection including WithinUsAI/Phi4-Reasoner-Uncensored-gguf

WithIn US AI (((GGUF MODELS)))

Collection

LLM MODELS TRAINED, FINE-TUNED, MERGED BY (WITHIN US AI) • 23 items • Updated 8 days ago • 8