Instructions to use deepsweet/Qwen3.6-27B-MLX-VL-oQ6 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deepsweet/Qwen3.6-27B-MLX-VL-oQ6 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("deepsweet/Qwen3.6-27B-MLX-VL-oQ6")

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 deepsweet/Qwen3.6-27B-MLX-VL-oQ6 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "deepsweet/Qwen3.6-27B-MLX-VL-oQ6"

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": "deepsweet/Qwen3.6-27B-MLX-VL-oQ6"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use deepsweet/Qwen3.6-27B-MLX-VL-oQ6 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 "deepsweet/Qwen3.6-27B-MLX-VL-oQ6"

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 deepsweet/Qwen3.6-27B-MLX-VL-oQ6

Run Hermes

hermes

MLX LM

How to use deepsweet/Qwen3.6-27B-MLX-VL-oQ6 with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "deepsweet/Qwen3.6-27B-MLX-VL-oQ6"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "deepsweet/Qwen3.6-27B-MLX-VL-oQ6"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "deepsweet/Qwen3.6-27B-MLX-VL-oQ6",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Version that preserves MTP Heads

by david-king-neuma - opened 21 days ago

Discussion

david-king-neuma

21 days ago

As of v0.3.9 oMLX supports native MTP generation. This is very helpful for the 27B dense model token generation. Any chance we can get a version that preserves the MTP headers.

Jundot has a version but the model card is very bare on details. (https://huggingface.co/Jundot/Qwen3.6-27B-oQ6-mtp). Do you think this is sufficient? I like that you have text and vision versions.

Additionally, I love your KL Divergence graphs and explanation of "fp16", thanks for that!!

deepsweet

Owner 21 days ago

Hi.

Jundot is the author of oMLX and oQ, his uploads are definitely trustworthy.

As for MTP – I can upload a text-only Qwen3.6-27B-MLX-oQ6-MTP if you need it.

david-king-neuma

19 days ago

Thanks! I don't want to bother you, I am fine using Jundot's! Appreciate your willingness.

david-king-neuma changed discussion status to closed 19 days ago

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