Instructions to use Jackrong/Qwen3.8-27B-MTP-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Jackrong/Qwen3.8-27B-MTP-GGUF with Transformers:

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

pipe = pipeline("image-text-to-text", model="Jackrong/Qwen3.8-27B-MTP-GGUF")
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 AutoModel
model = AutoModel.from_pretrained("Jackrong/Qwen3.8-27B-MTP-GGUF", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Jackrong/Qwen3.8-27B-MTP-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Jackrong/Qwen3.8-27B-MTP-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Jackrong/Qwen3.8-27B-MTP-GGUF",
		"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/Jackrong/Qwen3.8-27B-MTP-GGUF

SGLang

How to use Jackrong/Qwen3.8-27B-MTP-GGUF 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 "Jackrong/Qwen3.8-27B-MTP-GGUF" \
    --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": "Jackrong/Qwen3.8-27B-MTP-GGUF",
		"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 "Jackrong/Qwen3.8-27B-MTP-GGUF" \
        --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": "Jackrong/Qwen3.8-27B-MTP-GGUF",
		"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
Docker Model Runner
How to use Jackrong/Qwen3.8-27B-MTP-GGUF with Docker Model Runner:
```
docker model run hf.co/Jackrong/Qwen3.8-27B-MTP-GGUF
```

Any plans for QWOPUS 3.8 ?

by gohv123 - opened 28 days ago

Discussion

gohv123

28 days ago

Hey Jackrong,

Any plans to release a 3.8 Qwopus ?

mvalley

27 days ago

Please, Qwopus 3.8!

mrpfeiler

26 days ago

I second that! 🥺 Qwopus 3.6 was ( ...and still is!!) sooooo good, it was my daily driver since it came out and always delivered as expected on my 3090+4070ti+32GB RAM Frankenstein setup 👌EXCELLENT WORK!!! -> 🤝please🤝 create Qwopus3.8-27B-v2-MTP-Q6 🤝 and Qwopus3.8-27B-v2-MTP-Q4 🥺 👍 👍 👍

Jackrong

Owner 26 days ago

Hey Jackrong,

Any plans to release a 3.8 Qwopus ?

Please, Qwopus 3.8!

I second that! 🥺 Qwopus 3.6 was ( ...and still is!!) sooooo good, it was my daily driver since it came out and always delivered as expected on my 3090+4070ti+32GB RAM Frankenstein setup 👌EXCELLENT WORK!!! -> 🤝please🤝 create Qwopus3.8-27B-v2-MTP-Q6 🤝 and Qwopus3.8-27B-v2-MTP-Q4 🥺 👍 👍 👍

Hi everyone, and thank you so much for the support!

Qwen3.8‑27B has shown extremely strong capabilities in my recent tests — in fact, it already surpasses many closed‑source teacher models. Because of that, doing casual or aggressive fine‑tuning could easily damage the model’s original strengths, so I’ve been very careful with the process.

But don’t worry — the new Qwopus3.8 is already in training.
I’m planning to add some RL training afterward as well, hopefully improving the efficiency issues that Qwen models sometimes run into. This whole process will take a bit more time, but progress is steady.

Once there’s any update, Kyle will announce it on X so everyone can follow along.

After this round of fine‑tuning, I’ll also share all the SFT data, the RL repo, and a full explanation of the fine‑tuning process. I really hope more people can join the world of large‑model development and build the models they love.
Thanks again for all your enthusiasm and support.

Best
Jack

haibin0528

25 days ago

Looking forward to it!

nonubitta

25 days ago

•

edited 25 days ago

Please create qwen3.8-9b-claude-4.6-opus-reasoning-distilled,
I use qwen3.5-9b-claude-4.6-opus-reasoning-distilled and it is wonderful.
For some reason, this is the only model which works with vs code and flutter, edit files automatically.

timewithaz

24 days ago

waiting for Qwopus3.8-Coder, thank you, Jack.

rightguard

23 days ago

Hello Jack, can you distill from Fable5 to create a brand new model such as Qwfable 3.8? Of course, Qwopus 3.8 is also great!

guell00

23 days ago

Hey Jack! Huge fan of your work, really love what you've been doing on HF.

Since you asked for suggestions for the Qwen distillation pipeline, here is a crazy idea:

Instead of just generic reasoning/math, could we bake in heavy factual grounding and single-prompt creative front-end into the teacher prompts?

Factual & Historical Anchor Prompts: Force the model in the distillation prompt to cite specific dates, historical events, and verified facts step-by-step before answering. That way, the student model actually inherits solid factual recall instead of just stylistic reasoning.

Interactive Front-end (Three.js / Canvas in single prompt): Prompts designed to output complete, self-contained mini-games (like Flappy Bird, mini 3D maps, procedural terrain) in pure HTML/JS + Three.js, completely runnable in one shot.

Having a small distilled model that can one-shot interactive Three.js scenes while retaining deep factual knowledge would be absolutely insane.

Hope you consider this! Keep crushing it! 🔥

chukk

19 days ago

Please, Qwopus 3.8!

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