Instructions to use banburist/Qwen3.5-9b-4bit-MTPLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use banburist/Qwen3.5-9b-4bit-MTPLX with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("banburist/Qwen3.5-9b-4bit-MTPLX")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use banburist/Qwen3.5-9b-4bit-MTPLX with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "banburist/Qwen3.5-9b-4bit-MTPLX"

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "banburist/Qwen3.5-9b-4bit-MTPLX"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use banburist/Qwen3.5-9b-4bit-MTPLX with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "banburist/Qwen3.5-9b-4bit-MTPLX"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "banburist/Qwen3.5-9b-4bit-MTPLX"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "banburist/Qwen3.5-9b-4bit-MTPLX",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use banburist/Qwen3.5-9b-4bit-MTPLX with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "banburist/Qwen3.5-9b-4bit-MTPLX"

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 banburist/Qwen3.5-9b-4bit-MTPLX

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use banburist/Qwen3.5-9b-4bit-MTPLX with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "banburist/Qwen3.5-9b-4bit-MTPLX"

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "banburist/Qwen3.5-9b-4bit-MTPLX" \
  --custom-provider-id mlx-lm \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Qwen3.5-9b-4bit-MTPLX

This is a quantized version of Qwen/Qwen3.5-9B for Apple Silicon (MLX) utilizing Multi-Token Prediction via MTPLX.

Details

Original Model: Qwen/Qwen3.5-9B
Quantization: 4-bit
Group Size: 64
Framework: MTPLX

Verification Stats

Best depth: D2
Multiplier vs autoregressive baseline: 1.97×
Verified on: Apple M5 Pro
Sampler: temperature 0.6 · top_p 0.95 · top_k 20

See mtplx_runtime.json for the full verification record.

Conversion

Built with MTPLX Forge

Using the App

Grab the Latest Release and read the App Instructions

Using the CLI

Install MTPLX

# Homebrew
brew install youssofal/mtplx/mtplx
# Pip
python3 -m pip install mtplx

Pull down the model:

# MTPLX picks this model up automatically when downloaded
mtplx pull banburist/Qwen-Qwen3.5-9b-4bit-MTPLX

Tune the model:

# Tune immediately after it is pulled
mtplx tune --model <model-or-path> --retune

Start MTPLX:

# for start chat
mtplx start
# for API server only
mtplx serve --port 8000

See the MTPLX repo for more information on configuring the model.

About

This was built to meet the hardware limitations of a 16gb machine without sacrificing MTPLX's performance improvements.

Use Youssof Altouhki's original 6-bit model if you have the hardware: Youssofal/Qwen3.5-9B-MTPLX-Optimized-Speed

Credits

Visit Youssof Altoukhi's profile to see all the official MTPLX builds.

License

This model inherits the Apache 2.0 license from the original Qwen3.5-9B model.

Downloads last month: 95

Safetensors

Model size

1B params

Tensor type

BF16

U32

F32

MLX

Hardware compatibility

4-bit

Model tree for banburist/Qwen3.5-9b-4bit-MTPLX

Base model

Qwen/Qwen3.5-9B-Base

Finetuned

Qwen/Qwen3.5-9B

Quantized

(472)

this model