Instructions to use mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF",
	filename="Qwen3.6-35B-A3B-uncensored-heretic-APEX-Balanced.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 mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF
# Run inference directly in the terminal:
llama-cli -hf mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF
# Run inference directly in the terminal:
llama-cli -hf mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF

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 mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF
# Run inference directly in the terminal:
./llama-cli -hf mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF

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 mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF
# Run inference directly in the terminal:
./build/bin/llama-cli -hf mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF

Use Docker

docker model run hf.co/mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF

LM Studio
Jan
Ollama
How to use mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF with Ollama:
```
ollama run hf.co/mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF
```

Unsloth Studio

How to use mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-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 mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-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 mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF to start chatting

How to use mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF

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": "mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-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 mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF

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 mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF

Run Hermes

hermes

Docker Model Runner
How to use mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF with Docker Model Runner:
```
docker model run hf.co/mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF
```

Lemonade

How to use mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull mudler/Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF

Run and chat with the model

lemonade run user.Qwen3.6-35B-A3B-uncensored-heretic-APEX-GGUF-{{QUANT_TAG}}

List all available models

lemonade list

Proposal for the Ultimate Local SOTA:

by netforcetech - opened 27 days ago

Discussion

netforcetech

27 days ago

Could you create such a coding model ?.So that it specializes as much as possible in coding.

I propose a model that has not yet been created, designed for the capabilities of existing 2026 technology.
A project that only someone with your hardware cluster can execute.

The Target Model:

DeepSeek-V4-Code-PRO-35B-A3B-MOE-Claude-4.7-Opus-Distill-i1-APEX-Thinking-UD-GGUF

The Technical Execution Pipeline:

Teacher Model (The Brains): Use DeepSeek-V4-PRO (1.6T) and Claude 4.7 Opus in a dual-teacher setup. We want the PRO's raw coding power combined with Opus's agentic logic.

Architecture (The Body): Distill that massive knowledge into a 35B-A3B (Active 3B) MoE structure. This is the only way to get 20 tok/sec on 32GB RAM / 8GB/12GB VRAM systems.

Training Method (Unsloth): Use Unsloth’s April 2026 Agentic Kernels. We need the model to have 'UD' (Universal Deployment) capabilities and native Tool-Use.

Logic Guard (APEX): Use APEX during the distillation phase. We cannot afford to lose the 'Thinking' chain-of-thought logic when we compress the PRO knowledge into the 35B frame.

Final Compression (mradermacher i1): The end result must be an iMatrix GGUF (Q5_K_M) targeting exactly 24.7 GB on disk.

This isn't just another merge. This is a PRO-grade distillation for the local user. You have the H100s to make this happen. If you bake this, it will be the #1 coding model globally for local hardware.
Make history.

"P.S. The community with 32GB RAM is currently stuck with older 35B models. If you are the first to drop this PRO-distilled 35B-A3B beast, it will hit the #1 Trending spot on Hugging Face within 24 hours. The demand is massive."

“A model that understands all languages and all programming languages, nothing more — everything must fit into it.”

veldierin

27 days ago

Could you create such a coding model ?.So that it specializes as much as possible in coding.

I propose a model that has not yet been created, designed for the capabilities of existing 2026 technology.
A project that only someone with your hardware cluster can execute.

The Target Model:

DeepSeek-V4-Code-PRO-35B-A3B-MOE-Claude-4.7-Opus-Distill-i1-APEX-Thinking-UD-GGUF

https://huggingface.co/mradermacher/Q3.6-27B-DS-v4-Flash-DA-i1-GGUF
https://huggingface.co/prithivMLmods/Q3.6-27B-DS-v4-Flash-DA
you can take a link to the original model if you want, but that looks to be what you're looking for. i haven't tested it yet but i'll definitely be giving it a shot.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment