Instructions to use ManniX-ITA/Qwen3.6-27B-A3B-CoderX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ManniX-ITA/Qwen3.6-27B-A3B-CoderX with Transformers:

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

pipe = pipeline("text-generation", model="ManniX-ITA/Qwen3.6-27B-A3B-CoderX")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ManniX-ITA/Qwen3.6-27B-A3B-CoderX")
model = AutoModelForCausalLM.from_pretrained("ManniX-ITA/Qwen3.6-27B-A3B-CoderX", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ManniX-ITA/Qwen3.6-27B-A3B-CoderX with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ManniX-ITA/Qwen3.6-27B-A3B-CoderX"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ManniX-ITA/Qwen3.6-27B-A3B-CoderX",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ManniX-ITA/Qwen3.6-27B-A3B-CoderX

SGLang

How to use ManniX-ITA/Qwen3.6-27B-A3B-CoderX 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 "ManniX-ITA/Qwen3.6-27B-A3B-CoderX" \
    --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": "ManniX-ITA/Qwen3.6-27B-A3B-CoderX",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "ManniX-ITA/Qwen3.6-27B-A3B-CoderX" \
        --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": "ManniX-ITA/Qwen3.6-27B-A3B-CoderX",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ManniX-ITA/Qwen3.6-27B-A3B-CoderX with Docker Model Runner:
```
docker model run hf.co/ManniX-ITA/Qwen3.6-27B-A3B-CoderX
```

Best performing model of qwen3.6 family

by repla73 - opened 6 days ago

Discussion

repla73

6 days ago

It's the best perfoming agentic model in this family and can run 15+ tok/sec on modern cpu's

Thanks for this great work

joeblowma

about 15 hours ago

Finally got around to trying this one the other day while I was without internet (electrician contractors cut fiber by accident) and I have to agree, at least for the Qwen MoEs (I'm pretty fond of a couple of the extra trained 3.5-9B) this one performs really well. Getting over 30t/s with my RTX4060 even at 128K filled context. It did a solid one shot conversion of a complicated batch file to powershell, given some json data it had a couple errors to deal with but practically one shotted a python script to import csv data (hundreds of entries) into the json config.

Like all Qwens on the smaller side, it really does seem to do a lot of "no wait..." in it's thinking 😆

repla73

about 14 hours ago

Finally got around to trying this one the other day while I was without internet (electrician contractors cut fiber by accident) and I have to agree, at least for the Qwen MoEs (I'm pretty fond of a couple of the extra trained 3.5-9B) this one performs really well. Getting over 30t/s with my RTX4060 even at 128K filled context. It did a solid one shot conversion of a complicated batch file to powershell, given some json data it had a couple errors to deal with but practically one shotted a python script to import csv data (hundreds of entries) into the json config.

Like all Qwens on the smaller side, it really does seem to do a lot of "no wait..." in it's thinking 😆

"coder" version works better without even thinking on. You should try that if you want the faster and more accurate. I created a real-world benchmark and tested. This version thinks much longer, better for planning stuff, coder version is for more agentic tool usage.

joeblowma

about 13 hours ago

•

edited about 13 hours ago

Oh nice! I didn't even realize there was a coder and coderx separate version. I just finished pulling the Q6 weights of this to see if they do better than IQ4_XS... maybe I should have saved the space for the alt! 😆

edit:/ thanks for the heads-up and reply repla!

ManniX-ITA

Owner about 4 hours ago

"coder" version works better without even thinking on. You should try that if you want the faster and more accurate. I created a real-world benchmark and tested. This version thinks much longer, better for planning stuff, coder version is for more agentic tool usage.

They are different and indeed the coder version is much better at multi-turn agentic coding workload.
coderx is more "capable" but it thinks a lot more, about 3x, and it gets lost in long running tasks, it's better only on evaluations and one-shot complex tasks

The best model overall and specifically for multi-turn agentic coding it's still Omnimerge-v4 and by a great margin.
It crushes even the new version, Qwen 3.8 27B

repla73

about 3 hours ago

Thanks,
Actually I'm waiting for 3.8 version of Omniverge : ) Do you have plans for that? Please do!

ManniX-ITA

Owner about 3 hours ago

Thanks,
Actually I'm waiting for 3.8 version of Omniverge : ) Do you have plans for that? Please do!

Building it now, will be Omnimerge v6.

repla73

about 2 hours ago

Thanks! That's extremely exciting!

joeblowma

about 2 hours ago

The best model overall and specifically for multi-turn agentic coding it's still Omnimerge-v4 and by a great margin.
It crushes even the new version, Qwen 3.8 27B

I'm just happy there is a middle ground between 9B and 40B that I can actually run, getting to compare these models to frontier models is turning out to be a rather fun sidequest.

Personally I'm waiting to see how the hardware shakes out... what form non-gpu iterated hardware will take for large models and whether it will stick to the insane GPU pricing models.

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