Instructions to use jakejharris/jspark3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jakejharris/jspark3 with Transformers:

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

pipe = pipeline("image-text-to-text", model="jakejharris/jspark3")
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("jakejharris/jspark3")
model = AutoModelForMultimodalLM.from_pretrained("jakejharris/jspark3", 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 jakejharris/jspark3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jakejharris/jspark3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jakejharris/jspark3",
		"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/jakejharris/jspark3

SGLang

How to use jakejharris/jspark3 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 "jakejharris/jspark3" \
    --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": "jakejharris/jspark3",
		"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 "jakejharris/jspark3" \
        --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": "jakejharris/jspark3",
		"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 jakejharris/jspark3 with Docker Model Runner:
```
docker model run hf.co/jakejharris/jspark3
```

JSpark3 v1.8.3

A serving recipe for GLM-5.3 Flash on three NVIDIA DGX Sparks, with an attributed mirror of the target weights.

Current release: v1.8.3. Start with the v1.8.3 release and the v1.8.3 installation guide.

Do not install v1.8.0, including the old archive in this HF repo: it requires a container image that was never published.

Download the v1.8.3 recipe tarball from GitHub, extract it, and run the checksum and validator commands below from the extracted jspark3 directory. Or clone the v1.8.3 source tag:

Use a shallow clone: a full git clone fails v1.8.2's privacy scan on old commit history.

git clone --depth 1 --branch v1.8.3 https://github.com/jakejharris/jspark3.git
cd jspark3
sha256sum -c SHA256SUMS
python3 -B tools/validate_release.py .

Then follow the installation guide to build and verify your own local image and native binaries. There is no JSpark3 image to pull from GHCR. Keep the default JSPARK3_V16_COOP=0; cache preparation, operator hygiene and three-host hardware qualification remain required before serving traffic.

Weights and measurements

The mirrored target weights are unchanged. For serving, use Mia-AiLab's pinned weights or this mirror at its verified revision, as required by the recipe's checksum ledger. DFlash2 is a separate download described in the installation guide.

Historical measurements reached up to 141.7 tok/s code and 90.7 tok/s prose at four streams (best of two runs, stock weights). These unchanged v1.8.0 measurements used coop=1; they do not qualify or predict the default operator coop=0 configuration. See the measurement definitions and full ranges and frozen results.

Licenses and attribution

The target weights remain under the ShapleyMcg License v1.0 in LICENSE. Z.AI's base model is MIT. JSpark3's original code and prose are Apache-2.0; included derivatives retain AGPL-3.0-only and vendored headers retain MIT. DFlash2 retains its non-commercial research/evaluation restriction. See the v1.8.3 licensing guide for component boundaries and notices.

This work includes or was produced using ShapleyMcg, created by Brandon M. Music (https://github.com/brandonmmusic-max/shapleymcg). ShapleyMcg is licensed under the ShapleyMcg License v1.0, an attribution-required license that grants no rights to the person known as "0xSero." Use of ShapleyMcg without this attribution is unlicensed.

Downloads last month: 196

Safetensors

Model size

88B params

Tensor type

F32

I32

BF16

F16

I16

Model tree for jakejharris/jspark3

Base model

zai-org/GLM-5.3-Flash

Quantized

(124)

this model