Instructions to use matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit with MLX:

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

from mlx_vlm import load, generate
from mlx_vlm.prompt_utils import apply_chat_template
from mlx_vlm.utils import load_config

# Load the model
model, processor = load("matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit")
config = load_config("matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit")

# Prepare input
image = ["http://images.cocodataset.org/val2017/000000039769.jpg"]
prompt = "Describe this image."

# Apply chat template
formatted_prompt = apply_chat_template(
    processor, config, prompt, num_images=1
)

# Generate output
output = generate(model, processor, formatted_prompt, image)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps
LM Studio

Unsloth Studio new

How to use matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit 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 matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit 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 matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit",
    max_seq_length=2048,
)

Pi new

How to use matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit 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 "matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit"

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 matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit

Run Hermes

hermes

MLX-Qwopus3.5-27B-v3-vision-4bit

A 4-bit MLX quantization of Jackrong/Qwopus3.5-27B-v3 with a few tweaks to restore the multimodal capabilities.

Supports {%- set enable_thinking = false %} Jinja variable.

Update [2026-04-12]: Refined the chat template to further improve stability for long-running tasks and tool use; mitigated an issue where incorrect think tag formatting could leak from a distillation dataset.

Quantization Details

Property	Value
Method	4-bit (4.695 bits per weight)
Tool	`mlx-vlm 0.4.2` via `mlx-vlm.convert`
Size	~16.1GB

Other Available Quants

Model	Size	Quantization	Bits per weight	Multimodal
Jackrong/MLX-Qwopus3.5-27B-v3-4bit	15.15 GB	4-bit	4.501	✗
(This model)	16.08 GB	4-bit	4.695	✓ (Vision)
matt-here/MLX-Qwopus3.5-27B-v3-5bit	18.56 GB	5-bit	5.501	✗
matt-here/MLX-Qwopus3.5-27B-v3-vision-5bit	19.46 GB	5-bit	5.678	✓ (Vision)
Jackrong/MLX-Qwopus3.5-27B-v3-6bit	21.88 GB	6-bit	6.501	✗
matt-here/MLX-Qwopus3.5-27B-v3-vision-6bit	22.85 GB	6-bit	6.661	✓ (Vision)
Jackrong/MLX-Qwopus3.5-27B-v3-bf16	53.81 GB	bf16	16	✗

GGUF quants - Jackrong/Qwopus3.5-27B-v3-GGUF

Credits

Alibaba Qwen Team — Qwen 3.5 27B dense model
Jackrong - Claude 4.6 Opus v3 distillation work
Unsloth - Training framework
Apple MLX Team - High-speed local inference on Apple Silicon

Downloads last month: 45

Safetensors

Model size

5B params

Tensor type

BF16

U32

MLX

Hardware compatibility

4-bit

Model tree for matt-here/MLX-Qwopus3.5-27B-v3-vision-4bit

Base model

Qwen/Qwen3.5-27B

Finetuned

unsloth/Qwen3.5-27B

Adapter

Jackrong/Qwopus3.5-27B-v3

Adapter

(11)

this model