Instructions to use AstraiaAI/Qwen3.5-4B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AstraiaAI/Qwen3.5-4B-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="AstraiaAI/Qwen3.5-4B-GGUF",
	filename="Qwen3.5-4B-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 AstraiaAI/Qwen3.5-4B-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf AstraiaAI/Qwen3.5-4B-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 AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf AstraiaAI/Qwen3.5-4B-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 AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf AstraiaAI/Qwen3.5-4B-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 AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use AstraiaAI/Qwen3.5-4B-GGUF with Ollama:
```
ollama run hf.co/AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M
```

Unsloth Studio

How to use AstraiaAI/Qwen3.5-4B-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 AstraiaAI/Qwen3.5-4B-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 AstraiaAI/Qwen3.5-4B-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for AstraiaAI/Qwen3.5-4B-GGUF to start chatting

How to use AstraiaAI/Qwen3.5-4B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf AstraiaAI/Qwen3.5-4B-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": "AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use AstraiaAI/Qwen3.5-4B-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 AstraiaAI/Qwen3.5-4B-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 AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use AstraiaAI/Qwen3.5-4B-GGUF with Docker Model Runner:
```
docker model run hf.co/AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M
```

Lemonade

How to use AstraiaAI/Qwen3.5-4B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull AstraiaAI/Qwen3.5-4B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Qwen3.5-4B-GGUF-Q4_K_M

List all available models

lemonade list


░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒░░
░░▒▒▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▒▒░░
░░▒▒▓▓                                                          ▓▓▒▒░░
░░▒▒▓▓    .                    *            .          ·        ▓▓▒▒░░
░░▒▒▓▓                                                          ▓▓▒▒░░
░░▒▒▓▓   █████╗ ███████╗████████╗██████╗  █████╗ ██╗ █████╗     ▓▓▒▒░░
░░▒▒▓▓  ██╔══██╗██╔════╝╚══██╔══╝██╔══██╗██╔══██╗██║██╔══██╗    ▓▓▒▒░░
░░▒▒▓▓  ███████║███████╗   ██║   ██████╔╝███████║██║███████║    ▓▓▒▒░░
░░▒▒▓▓  ██╔══██║╚════██║   ██║   ██╔══██╗██╔══██║██║██╔══██║    ▓▓▒▒░░
░░▒▒▓▓  ██║  ██║███████║   ██║   ██║  ██║██║  ██║██║██║  ██║    ▓▓▒▒░░
░░▒▒▓▓  ╚═╝  ╚═╝╚══════╝   ╚═╝   ╚═╝  ╚═╝╚═╝  ╚═╝╚═╝╚═╝  ╚═╝    ▓▓▒▒░░
░░▒▒▓▓                                                          ▓▓▒▒░░
░░▒▒▓▓                     -  MMXXVI  -                         ▓▓▒▒░░
░░▒▒▓▓                                                          ▓▓▒▒░░
░░▒▒▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▒▒░░
░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒░░
░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

╔══════════════════════════════════════════════════════════════════════╗
║                                                                      ║
║      ░▒▓█  Q W E N 3 . 5 - 4 B  ·  G G U F  R E L E A S E  █▓▒░      ║
║                                                                      ║
║   ──────────────────────────────────────────────────────────────     ║
║                a   s   t   r   a   i   a       a   i                 ║
║   ──────────────────────────────────────────────────────────────     ║
║                                                                      ║
║         release ...: Qwen3.5-4B-GGUF                                 ║
║         format ....: GGUF (llama.cpp / ggml)                         ║
║         params ....: ~4.0B                                           ║
║         base model : Qwen/Qwen3.5-4B                                 ║
║         Improvised by ....: AstraiaAI                                ║
║         size .pkg .: 21.9 GB total                                   ║
║                                                                      ║
╚══════════════════════════════════════════════════════════════════════╝

Overall Model Intelligence Score: 0.765 Stirlings (Equivalent to roughly 76% of human expert performance). Complete report: Engram.md

   .------------------------------------------------------------------.
  /  > 0x01   A B O U T   T H I S   R E L E A S E                      \
 /____________________________________________________________________/

These are GGUF quantizations of Qwen3.5-4B, intended for fast local inference with llama.cpp, LM Studio, Ollama, text-generation-webui, koboldcpp, and anything else that speaks ggml.

Two flavors ship in this repo:

   ┌───────────────────────────┬────────────────────────────────────────┐
   │  FLAVOR                   │  BEHAVIOR                              │
   ├───────────────────────────┼────────────────────────────────────────┤
   │  Qwen3.5-4B               │   stock build — reasoning / thinking   │
   │                           │   traces enabled per upstream config   │
   ├───────────────────────────┼────────────────────────────────────────┤
   │  _NO_REASONING_DEFAULTS   │   same weights, baked-in defaults set  │
   │                           │   to suppress chain-of-thought output  │
   └───────────────────────────┴────────────────────────────────────────┘

Pick the plain build if you want the model to "think" before answering. Pick NO_REASONING_DEFAULTS if you want terse, direct replies out of the box without flipping flags every session.

 ┌────────────────────────────────────────────────────────┬─────────┬──────┐
 │  FILE                                                  │  SIZE   │ BITS │
 ├────────────────────────────────────────────────────────┼─────────┼──────┤
 │  Qwen3.5-4B-Q4_K_M.gguf                                │ 2.78 GB │ ~4   │
 │  Qwen3.5-4B-Q6_K.gguf                                  │ 3.56 GB │ ~6   │
 │  Qwen3.5-4B-Q8_0.gguf                                  │ 4.61 GB │  8   │
 ├────────────────────────────────────────────────────────┼─────────┼──────┤
 │  Qwen3.5-4B_NO_REASONING_DEFAULTS-Q4_K_M.gguf          │ 2.78 GB │ ~4   │
 │  Qwen3.5-4B_NO_REASONING_DEFAULTS-Q6_K.gguf            │ 3.56 GB │ ~6   │
 │  Qwen3.5-4B_NO_REASONING_DEFAULTS-Q8_0.gguf            │ 4.61 GB │  8   │
 └────────────────────────────────────────────────────────┴─────────┴──────┘
                                              total ......: 21.9 GB

   .------------------------------------------------------------------.
  /  > 0x02   Q U A N T   S E L E C T I O N   G U I D E                \
 /____________________________________________________________________/

   Q4_K_M   ███████░░░   smallest / fastest. best for tight VRAM or RAM.
            quality is acceptable; some degradation present vs higher bits.

   Q6_K     █████████░   sweet spot. near-fp16 quality, modest footprint.
            recommended default for most users.

   Q8_0     ██████████   Indistinguishable from full-precision. heaviest of the
            three. use when you want maximum accuracy and have strong hardware.

            ╔══════════════════════════════════════════════════╗
            ║   not sure? → grab Q6_K and move on with life.   ║
            ╚══════════════════════════════════════════════════╝

   .------------------------------------------------------------------.
  /  > 0x03   Q U I C K   S T A R T                                    \
 /____________________________________________________________________/

llama.cpp (CLI)

# download a single quant
huggingface-cli download AstraiaAI/Qwen3.5-4B-GGUF \
    Qwen3.5-4B-Q6_K.gguf --local-dir . --local-dir-use-symlinks False

# run it
./llama-cli -m Qwen3.5-4B-Q6_K.gguf -p "Explain quantization in detail." -n 256

llama.cpp (OpenAI-compatible server)

./llama-server -m Qwen3.5-4B-Q6_K.gguf -c 8192 --port 8080
# → http://localhost:8080

Ollama

# Modelfile
printf 'FROM ./Qwen3.5-4B-Q6_K.gguf\n' > Modelfile
ollama create qwen3.5-4b -f Modelfile
ollama run qwen3.5-4b

LM Studio / koboldcpp / text-generation-webui

Drop any .gguf above into your models folder and load it from the UI.

   .------------------------------------------------------------------.
  /  > 0x04   R E A S O N I N G   N O T E S                            \
 /____________________________________________________________________/

   ┌─[ stock build ]───────────────────────────────────────────────┐
   │  emits thinking / reasoning content before the final answer.  │
   │  great for math, code, multi-step problems. slower per reply. │
   └───────────────────────────────────────────────────────────────┘

   ┌─[ NO_REASONING_DEFAULTS build ]───────────────────────────────┐
   │  defaults tuned to skip the chain-of-thought and answer       │
   │  directly. faster, cleaner for chat / RAG / agents that don't │
   │  want to parse <think> blocks. same underlying weights.       │
   └───────────────────────────────────────────────────────────────┘

   .------------------------------------------------------------------.
  /  > 0x05   L I C E N S E   &   C R E D I T S                        \
 /____________________________________________________________________/

   base weights ....: Qwen Team  (Qwen3.5-4B)
   quantization ....: AstraiaAI
   tooling .........: llama.cpp / ggml
   license .........: Apache-2.0  (inherits from base model)

Use of these weights is subject to the upstream Qwen license terms. These GGUF files are conversions only; no weights were retrained or fine-tuned.

 ┌──────────────────────────────────────────────────────────────────────┐
 │                                                                      │
 │         Salutes to everyone running 4B models on a potato   ☕       │
 │                                                                      │
 │              ░▒▓  A S T R A I A   A I  ·  MMXXVI  ▓▒░                │
 │                                                                      │
 └──────────────────────────────────────────────────────────────────────┘

Downloads last month: 699

GGUF

Model size

4B params

Architecture

qwen35

Hardware compatibility

4-bit

6-bit

8-bit

Inference Providers NEW

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