Instructions to use b1intern/video-9b-checkpoint-114 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use b1intern/video-9b-checkpoint-114 with Transformers:

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

pipe = pipeline("image-text-to-text", model="b1intern/video-9b-checkpoint-114")
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("b1intern/video-9b-checkpoint-114")
model = AutoModelForMultimodalLM.from_pretrained("b1intern/video-9b-checkpoint-114", 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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use b1intern/video-9b-checkpoint-114 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "b1intern/video-9b-checkpoint-114"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "b1intern/video-9b-checkpoint-114",
		"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/b1intern/video-9b-checkpoint-114

SGLang

How to use b1intern/video-9b-checkpoint-114 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 "b1intern/video-9b-checkpoint-114" \
    --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": "b1intern/video-9b-checkpoint-114",
		"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 "b1intern/video-9b-checkpoint-114" \
        --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": "b1intern/video-9b-checkpoint-114",
		"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 b1intern/video-9b-checkpoint-114 with Docker Model Runner:
```
docker model run hf.co/b1intern/video-9b-checkpoint-114
```

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Video 9B — Checkpoint 114

Hugging Face model-loading export of SFT checkpoint 114 from the run video_v2qa16418_9b_32gpu_gbs64_lr1e5_20261008T000457CST.

The repository contains four BF16 safetensors weight shards, their index, model configuration, tokenizer, processor configuration, and chat template. The model architecture is Qwen3_5ForConditionalGeneration.

Access and download

The repository is publicly visible. Request access on the repository page; requests are approved automatically. Download using your own approved account:

hf auth login
hf download b1intern/video-9b-checkpoint-114 --local-dir ./video-9b-checkpoint-114
cd video-9b-checkpoint-114
sha256sum -c SHA256SUMS

Use the downloaded directory as the model path with a compatible Qwen3.5 runtime and the matching video/tool inference harness. This export can be used for inference or initialization of a new RL run. Distributed SFT optimizer, RNG, scheduler, and trainer state are not included, so it is not an exact SFT training-state resume package.

All model file sizes and hashes, all 760 tensor entries, and the shard index are checked during publication. This transfer does not constitute a new model quality evaluation or a new GPU training/inference test.

Evaluation package

The evaluation directory provides the 2026-10-14 code handoff with both 64-frame and 2 FPS modes, integrated fixes and tests, plus an optional 5.04 GB video runtime archive. The default suite is six benchmarks without LongVideo-Reason. The code and runtime are separate from the model files.

For a weights-only download, add --exclude "evaluation/*" to the download command.

Video SFT pipeline package (2026-10-14)

The SFT pipeline handoff contains generation, review, reflection, export and training-format verification code. Start with download instructions, README, and Chinese quickstart. Data, videos, model assets and private provider clients are not included in this package; configure your own inputs and authorized provider integration.

Download only this package:

hf download b1intern/video-9b-checkpoint-114 --repo-type model --include "sft_pipeline/20261014/*" --local-dir ./video-sft-pipeline

Downloads last month: 3

Safetensors

Model size

9B params

Tensor type

BF16