Instructions to use NyxKrage/moondream3-preview-hf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use NyxKrage/moondream3-preview-hf with Transformers:

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

pipe = pipeline("image-text-to-text", model="NyxKrage/moondream3-preview-hf", trust_remote_code=True)
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 Moondream3ForConditonalGeneration
model = Moondream3ForConditonalGeneration.from_pretrained("NyxKrage/moondream3-preview-hf", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use NyxKrage/moondream3-preview-hf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "NyxKrage/moondream3-preview-hf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NyxKrage/moondream3-preview-hf",
		"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/NyxKrage/moondream3-preview-hf

SGLang

How to use NyxKrage/moondream3-preview-hf 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 "NyxKrage/moondream3-preview-hf" \
    --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": "NyxKrage/moondream3-preview-hf",
		"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 "NyxKrage/moondream3-preview-hf" \
        --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": "NyxKrage/moondream3-preview-hf",
		"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 NyxKrage/moondream3-preview-hf with Docker Model Runner:
```
docker model run hf.co/NyxKrage/moondream3-preview-hf
```

NyxKrage commited on Nov 26, 2025

Commit

1101625

verified ·

1 Parent(s): 2fa4720

fix bounding box detection not accounting for batch dimension

Browse files

Files changed (1) hide show

modeling_moondream3.py +7 -4

modeling_moondream3.py CHANGED Viewed

@@ -1147,12 +1147,15 @@ class Moondream3ForConditionalGeneration(Moondream3PreTrainedModel, GenerationMi
                         torch.stack([w, h],dim=-1).to(size_logits.dtype)
                     )
                 ).unsqueeze(1)
                 bbox = [
-                    x_center.item() - w.item() / 2,
-                    y_center.item() - h.item() / 2,
-                    x_center.item() + w.item() / 2,
-                    y_center.item() + h.item() / 2,
                 ]
                 bbox = bbox * (batch_mask).unsqueeze(1)
                 pos_ids += 1
                 cache_pos = cache_pos + 1

                         torch.stack([w, h],dim=-1).to(size_logits.dtype)
                     )
                 ).unsqueeze(1)
+                x_center = x_center.squeeze(1)
+                y_center = y_center.squeeze(1)
                 bbox = [
+                    x_center - w / 2,
+                    y_center - h / 2,
+                    x_center + w / 2,
+                    y_center + h / 2,
                 ]
+                bbox = torch.stack(bbox, dim=1)
                 bbox = bbox * (batch_mask).unsqueeze(1)
                 pos_ids += 1
                 cache_pos = cache_pos + 1