Instructions to use AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4 with Transformers:

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

pipe = pipeline("text-generation", model="AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4")
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("AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4")
model = AutoModelForMultimodalLM.from_pretrained("AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4")
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 AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4

SGLang

How to use AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4 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 "AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4" \
    --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": "AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4",
		"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 "AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4" \
        --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": "AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4 with Docker Model Runner:
```
docker model run hf.co/AEON-7/Qwen3.6-27B-AEON-Ultimate-Uncensored-NVFP4
```

AEON-7 commited on 11 days ago

Commit

c2ebb91

verified ·

1 Parent(s): fd7c7da

docs: correct DFlash KV cache compat — needs BF16 in vLLM 0.22.1 (regression vs v3)

Browse files

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -63,7 +63,7 @@ tags:
 >
 > Quick start + full DFlash recipe: [container README](https://github.com/AEON-7/vllm-ultimate-dgx-spark).
 >
-> ⚠️ **NVFP4 KV + DFlash is not yet compatible on sm_121a in vLLM 0.22.1**: the DFlash drafter uses non-causal attention, which currently has no NVFP4-KV backend on Spark (FLASHINFER requires SM100; TRITON_ATTN is causal-only). For DFlash use `--kv-cache-dtype fp8_e4m3`. NVFP4 KV pairs cleanly with MTP / Eagle / ngram speculators.
 ## Variants

 >
 > Quick start + full DFlash recipe: [container README](https://github.com/AEON-7/vllm-ultimate-dgx-spark).
 >
+> ⚠️ **DFlash in vLLM 0.22.1 needs `--kv-cache-dtype auto` (BF16) on Spark.** The DFlash drafter uses non-causal attention. In PR #44389's refactor, neither FLASHINFER nor TRITON_ATTN has a non-causal kernel for FP8 or NVFP4 KV anymore — only FLASH_ATTN works, and it's BF16-only. This is a regression vs the v3 production image (vLLM 0.20.0 had a FLASHINFER non-causal+FP8 path). For FP8-KV-with-DFlash today, **stay on v3**. For NVFP4 KV's ~3× capacity gain, use a causal speculator (MTP / Eagle / ngram) with the MTP-XS body.
 ## Variants