Instructions to use orcarouter/GLM-5.3-Flash-Uncensored-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use orcarouter/GLM-5.3-Flash-Uncensored-NVFP4 with Transformers:

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

pipe = pipeline("image-text-to-text", model="orcarouter/GLM-5.3-Flash-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("orcarouter/GLM-5.3-Flash-Uncensored-NVFP4")
model = AutoModelForMultimodalLM.from_pretrained("orcarouter/GLM-5.3-Flash-Uncensored-NVFP4", 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 orcarouter/GLM-5.3-Flash-Uncensored-NVFP4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "orcarouter/GLM-5.3-Flash-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": "orcarouter/GLM-5.3-Flash-Uncensored-NVFP4",
		"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/orcarouter/GLM-5.3-Flash-Uncensored-NVFP4

SGLang

How to use orcarouter/GLM-5.3-Flash-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 "orcarouter/GLM-5.3-Flash-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": "orcarouter/GLM-5.3-Flash-Uncensored-NVFP4",
		"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 "orcarouter/GLM-5.3-Flash-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": "orcarouter/GLM-5.3-Flash-Uncensored-NVFP4",
		"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 orcarouter/GLM-5.3-Flash-Uncensored-NVFP4 with Docker Model Runner:
```
docker model run hf.co/orcarouter/GLM-5.3-Flash-Uncensored-NVFP4
```

MTP drafter missing?

by kamikode - opened 3 days ago

Discussion

kamikode

3 days ago

The MTP drafter seems to have been dropped from the checkpoint. Is this on purpose?

randomly-selected1

3 days ago

Yep couldnt deploy the model with MTP. Also the latest layer - 45th glm_next - is missing

kamikode

3 days ago

Yes, actually after testing, the model seems to be wholly broken. It's just outputting "!!!!..." indefinitely

mmssix

2 days ago

Agreed. No MTP and will not serve. I think this might just be an email harvesting trap.

Per Opus:
glm-5.3-flash-uncensored-nvfp4 — cannot work on this host

Three attempts, all dead ends:

jovian image, as configured → ValidationError: checkpoint is compressed-tensors, launcher hardcodes --quantization modelopt_mixed.
- --quantization compressed-tensors → starts, serves 200 OK, but emits pure ! and logprobs return nan. Reproduced on two MoE kernels (B12X and FLASHINFER_CUTLASS). Root cause is the checkpoint: OrcaSAQ gave
  gate_proj and up_proj separate FP32 global scales (21504 vs 26496), and vLLM's fused w13 path keeps only gate's.
the model card's own image (pulled it) → pe_dim must be 64 for fp8_ds_mla. platforms/cuda.py gives SM120 only [TRITON_MLA, FLASHINFER_MLA_SPARSE_SM120], the latter mandates fp8_ds_mla, and GLM is NoPE
(qk_rope_head_dim=0). The SM90 branch has an explicit case for this shape; SM120 has none. The card's verified platform is 8×H100.

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