Instructions to use lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL with Transformers:

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

pipe = pipeline("image-text-to-text", model="lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL")
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("lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL")
model = AutoModelForMultimodalLM.from_pretrained("lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL", 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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL",
		"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/lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL

SGLang

How to use lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL 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 "lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL" \
    --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": "lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL",
		"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 "lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL" \
        --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": "lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL",
		"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"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL with Docker Model Runner:
```
docker model run hf.co/lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL
```

Huihui Qwen3.8-27B Abliterated NVFP4 MTP VL

A compressed-tensors NVFP4 W4A4 release of huihui-ai/Huihui-Qwen3.8-27B-abliterated at revision d42ca8978c5a66e92c3446d46e8adfe03ef692ff. The same-source BF16 vision/video tower (333 tensors) and all 15 BF16 MTP tensors are retained.

Quick start — RTX 5090 / Blackwell

hf download lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL --local-dir ./huihui-qwen38-nvfp4

docker run --rm --gpus all --ipc=host --network=host   -e VLLM_NVFP4_GEMM_BACKEND=flashinfer-cutlass   -e VLLM_USE_FLASHINFER_SAMPLER=1   -v "$PWD/huihui-qwen38-nvfp4:/model:ro"   vllm/vllm-openai:qwen38-x86_64-cu130   /model   --served-model-name qwen38-huihui-nvfp4   --host 0.0.0.0 --port 8000   --max-model-len 4096   --kv-cache-dtype fp8   --gpu-memory-utilization 0.92   --max-num-seqs 1   --max-num-batched-tokens 1024   --enable-prefix-caching   --trust-remote-code   --speculative-config '{"method":"qwen3_5_mtp","num_speculative_tokens":3}'

The command above is the verified multimodal/MTP smoke profile. Larger context values require workload-specific VRAM validation.

Quantization and lineage

Component	Format/source
Language-model Linear layers	NVFP4 W4A4, group size 16
Vision/video tower	333 tensors, BF16, same Huihui checkpoint
MTP head	15 tensors, BF16, same Huihui checkpoint
`lm_head`, token embedding, GDN `conv1d`	BF16
Calibration	CNN/DailyMail 3.0.0, 20 × 8192 tokens
Packaging	compressed-tensors `nvfp4-pack-quantized`

No official Qwen, Unsloth, Blackfrost, or other behavioral variant weights were grafted into this release.

Validation

Validated on one RTX 5090, 450 W cap, using vllm/vllm-openai:qwen38-x86_64-cu130:

GET /health and /v1/models: passed
OpenAI-compatible 1024-token text generation: passed
Native MTP n=3: passed
MTP draft tokens: 1362; accepted: 572; acceptance: 42.0%
Mean acceptance length: 2.33
Per-position acceptance: 0.602 / 0.429 / 0.295
Client elapsed for 1024 output tokens: 10.73 s
Runtime VRAM under generation: ~28,944 MiB

The earlier full release workflow also verifies 333 vision tensors and 15 same-source MTP tensors statically. The NInfer derivative in the companion repository carries a separate real-image runtime validation.

Files

model-00001-of-00002.safetensors, model-00002-of-00002.safetensors: compressed checkpoint
model-mtp-extra.safetensors: 15 same-source BF16 MTP tensors
model.safetensors.index.json: complete 2687-tensor index
BUILD_MANIFEST.json, VALIDATION_REPORT.json, STATIC_VALIDATION_REPORT.json, recipe.yaml, SHA256SUMS: provenance and reproducibility

Intended use

This is an abliterated behavioral derivative intended for model research and local inference. Users are responsible for downstream use and applicable policies.

Downloads last month: 747

Safetensors

Model size

28B params

Tensor type

BF16

Model tree for lyf/Qwen3.8-27B-Huihui-Abliterated-NVFP4-MTP-VL

Base model

Qwen/Qwen3.8-27B

Finetuned

huihui-ai/Huihui-Qwen3.8-27B-abliterated

Quantized

(48)

this model