Instructions to use nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451")
model = AutoModelForMultimodalLM.from_pretrained("nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

MLX

How to use nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 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("nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451")
config = load_config("nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451")

# 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 Settings
LM Studio

vLLM

How to use nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451

SGLang

How to use nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Unsloth Desktop

How to use nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451"

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": "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 with Docker Model Runner:
```
docker model run hf.co/nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451
```

Hermes Agent

How to use nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 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 "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451"

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 nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451 with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451"

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 "nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451" \
  --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"

Thoughtful and flexible

by meddle86 - opened Jul 19

Discussion

meddle86

Jul 19

Passed every real-world thing I've put it through thus far. Thinks and talks more like Opus than the vanilla Qwen version, and I think more abbreviated in its thinking (in a good way). Rigorous in attacking coding projects, kind of like GPT-5.6 if you've used it. I really like it. Nice work!

nightmedia

Owner Jul 19

Thank you very much for the feedback!

It does have quite a strong Opus base created by DavidAU, I built ontop of it. It also has Polaris Alpha(GPT) traces that define some of the behavior. The added books were to make the model language a bit more flexible and add some jokes :)

Zeaxion

Jul 25

Thank you very much for the feedback!

It does have quite a strong Opus base created by DavidAU, I built ontop of it. It also has Polaris Alpha(GPT) traces that define some of the behavior. The added books were to make the model language a bit more flexible and add some jokes :)

can you make it to FP8, FP16 is too big and speed slow for my device.Thanks

meddle86

Jul 27

•

edited Jul 27

I've done more extensive testing, and benched it against some other merges, as well as the original Qwen3.6-27b itself. I thought perhaps you may be interested, nightmedia: https://launchswitch.github.io/localvibebench/, https://launchswitch.github.io/localvibebench/writing/qwen36-27b-fable-fusion-711-benched/

I have lots of thinking traces and evidence that backs up the conclusions, and am happy to share.

In essence, I found this model is not quite as good of an open-book coder in my private codebases as vanilla Qwen3.6-27b, but thinks 30% more efficiently, which fits with my own experience outside of the benchmark as well. The primary weakness: verification and overconfidence (which also leads to the more efficient thinking: less questioning itself).

I also noticed the "Architect"play out in the thinking traces: this model helps make up for it's verification weakness with excellent "architecting" upfront.

I want to note that, so far, no other Qwen3.6-27b "blends", such as Tess and Qwopus Coder, have benched as highly as this one.

nightmedia

Owner Jul 27

Excellent review, and I thank you for that!

I embedded Tess as well, here is the mxfp4, mxfp8 and source are also available:
https://huggingface.co/nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451-Tess-mxfp4-mlx

This would be the 712 :)

You are correct to point out the overconfidence: the model was built that way.

Let me expand a bit on that. I do my merges on instrumentation only. I never changed the test prompt or the Genesis prompt since I started, and I tune models on specific behavioral patterns, one of which is confidence. Sometimes there can be a bit too much of it, which is why this model can do a pretty convincing Gul Dukat

https://huggingface.co/nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451-Tess-mxfp8-mlx

I do no coding or agentic tests, always use the same boring template, make as few changes to the structure as possible. If I merge a coding model, it will get better coding skills, with an author model it gets a better voice, and the more these are combined in the right order, you get good RP and coding in the same time.

The vibe at the end of confirming the numbers and the behavioral patterns, is what goes on the model card: it is always the first vibe, and not a regen.

If the model made a first impression, it goes up, otherwise it gets measured some more.

meddle86

Jul 27

Thank you for the background, very interesting!

I'm really happy to see this blow up like it has. Great to see yours and David's creativity rewarded.

Inspirational stuff!

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