Instructions to use TeichAI/Qwen3.6-27B-Fable-5-Experimental with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TeichAI/Qwen3.6-27B-Fable-5-Experimental with Transformers:

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

pipe = pipeline("image-text-to-text", model="TeichAI/Qwen3.6-27B-Fable-5-Experimental")
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("TeichAI/Qwen3.6-27B-Fable-5-Experimental")
model = AutoModelForMultimodalLM.from_pretrained("TeichAI/Qwen3.6-27B-Fable-5-Experimental")
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use TeichAI/Qwen3.6-27B-Fable-5-Experimental with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TeichAI/Qwen3.6-27B-Fable-5-Experimental"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TeichAI/Qwen3.6-27B-Fable-5-Experimental",
		"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/TeichAI/Qwen3.6-27B-Fable-5-Experimental

SGLang

How to use TeichAI/Qwen3.6-27B-Fable-5-Experimental 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 "TeichAI/Qwen3.6-27B-Fable-5-Experimental" \
    --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": "TeichAI/Qwen3.6-27B-Fable-5-Experimental",
		"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 "TeichAI/Qwen3.6-27B-Fable-5-Experimental" \
        --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": "TeichAI/Qwen3.6-27B-Fable-5-Experimental",
		"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 Studio

How to use TeichAI/Qwen3.6-27B-Fable-5-Experimental 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 TeichAI/Qwen3.6-27B-Fable-5-Experimental 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 TeichAI/Qwen3.6-27B-Fable-5-Experimental to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for TeichAI/Qwen3.6-27B-Fable-5-Experimental to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="TeichAI/Qwen3.6-27B-Fable-5-Experimental",
    max_seq_length=2048,
)

Docker Model Runner
How to use TeichAI/Qwen3.6-27B-Fable-5-Experimental with Docker Model Runner:
```
docker model run hf.co/TeichAI/Qwen3.6-27B-Fable-5-Experimental
```

Can this model be improved through RLAIF?

by FlameF0X - opened 2 days ago

Discussion

FlameF0X

2 days ago

First of all, congratulations for making this model. Sounds Interesting to play with it.

Second of all, can't be used to reward LM in fine tuning via RL? Like, take this model and try to find tune it using RLAIF where the feedback comes from the same model. Maybe it can make the model still could and feel like fable while making it smart?

Now I'm just hypnotising since I don't have any idea how long or expensive it would be.

(Sorry for my bad English)

FlameF0X changed discussion title from Can this model be user for RLAIF? to Can this model be improved through RLAIF? 2 days ago

armand0e

TeichAI org 2 days ago

No worries! I was actually planning on using the Fable 5 data to do SDFT (Self-Distillation Fine Tuning) which is a very similar concept. I don't really know what RLAIF is, but I'm currently doing a test run with the 9B to find some good parameters before committing to a more expensive run of SDFT.

More info on SDFT: https://arxiv.org/pdf/2601.19897

I do think certain aspects of it can be used for reward, maybe not for code quality, but for implementation planning and things like that definitely.

FlameF0X

2 days ago

RLAIF is RL from AI Feedback (https://arxiv.org/abs/2309.00267), it's like RLHF but ai does the whole work.

Also I never heard of SDFT, but sounds neat, I'm sure I'm going to check it out.

armand0e

TeichAI org 2 days ago

And I will do the same for RLAIF! Cheers :)

ianncity

1 day ago

First of all, congratulations for making this model. Sounds Interesting to play with it.

Second of all, can't be used to reward LM in fine tuning via RL? Like, take this model and try to find tune it using RLAIF where the feedback comes from the same model. Maybe it can make the model still could and feel like fable while making it smart?

Now I'm just hypnotising since I don't have any idea how long or expensive it would be.

(Sorry for my bad English)

Any type of RL is compute expensive far more expensive than sft, but RL is how models can really improve themselves. Reinforcement learnings only limit is the amount of money and time you have

FlameF0X

1 day ago

I know that money are a constraint for RL. I was just suggesting for the far future.

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