Instructions to use OpenGVLab/InternVL3_5-4B-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use OpenGVLab/InternVL3_5-4B-Instruct with Transformers:

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

pipe = pipeline("image-text-to-text", model="OpenGVLab/InternVL3_5-4B-Instruct", 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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("OpenGVLab/InternVL3_5-4B-Instruct", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use OpenGVLab/InternVL3_5-4B-Instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "OpenGVLab/InternVL3_5-4B-Instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "OpenGVLab/InternVL3_5-4B-Instruct",
		"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/OpenGVLab/InternVL3_5-4B-Instruct

SGLang

How to use OpenGVLab/InternVL3_5-4B-Instruct 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 "OpenGVLab/InternVL3_5-4B-Instruct" \
    --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": "OpenGVLab/InternVL3_5-4B-Instruct",
		"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 "OpenGVLab/InternVL3_5-4B-Instruct" \
        --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": "OpenGVLab/InternVL3_5-4B-Instruct",
		"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 OpenGVLab/InternVL3_5-4B-Instruct with Docker Model Runner:
```
docker model run hf.co/OpenGVLab/InternVL3_5-4B-Instruct
```

jazeelmohd commited on about 9 hours ago

Commit

48aeac9

verified ·

1 Parent(s): a3fd315

Add LongShOTBench evaluation results

Browse files

Add the LongShOTBench (NeurIPS 2026 Evaluations & Datasets Track) benchmark authors' published results for InternVL3_5-4B-Instruct on overall, core_perception, reasoning, information and multimodal. Each value is a rubric-weighted score on a 0-100 scale; higher is better.

LongShOTBench evaluates long-form omni-modal video understanding: 3,401 open-ended single- and multi-turn samples (4,882 QA turns) over 274 real-world videos (~41 min avg) that require combining visual, speech and ambient-audio evidence. Answers are graded by an LLM judge (Qwen3-14B) against human-verified, weighted criterion-level rubrics.

Source: https://longshot.cvmbzuai.com/leaderboard
Benchmark: https://huggingface.co/datasets/MBZUAI/longshot-bench
Paper: https://arxiv.org/abs/2512.16978

Files changed (1) hide show

.eval_results/longshotbench.yaml +45 -0

.eval_results/longshotbench.yaml ADDED Viewed

	@@ -0,0 +1,45 @@

+- dataset:
+    id: MBZUAI/longshot-bench
+    task_id: overall
+  value: 22.2
+  source:
+    url: https://longshot.cvmbzuai.com/leaderboard
+    name: LongShOTBench official results
+    user: MBZUAI
+  notes: 'Rubric-weighted score (0-100), LLM judge Qwen3-14B; model input: Video.'
+- dataset:
+    id: MBZUAI/longshot-bench
+    task_id: core_perception
+  value: 19.02
+  source:
+    url: https://longshot.cvmbzuai.com/leaderboard
+    name: LongShOTBench official results
+    user: MBZUAI
+  notes: 'Rubric-weighted score (0-100), LLM judge Qwen3-14B; model input: Video.'
+- dataset:
+    id: MBZUAI/longshot-bench
+    task_id: reasoning
+  value: 24.93
+  source:
+    url: https://longshot.cvmbzuai.com/leaderboard
+    name: LongShOTBench official results
+    user: MBZUAI
+  notes: 'Rubric-weighted score (0-100), LLM judge Qwen3-14B; model input: Video.'
+- dataset:
+    id: MBZUAI/longshot-bench
+    task_id: information
+  value: 19.58
+  source:
+    url: https://longshot.cvmbzuai.com/leaderboard
+    name: LongShOTBench official results
+    user: MBZUAI
+  notes: 'Rubric-weighted score (0-100), LLM judge Qwen3-14B; model input: Video.'
+- dataset:
+    id: MBZUAI/longshot-bench
+    task_id: multimodal
+  value: 25.25
+  source:
+    url: https://longshot.cvmbzuai.com/leaderboard
+    name: LongShOTBench official results
+    user: MBZUAI
+  notes: 'Rubric-weighted score (0-100), LLM judge Qwen3-14B; model input: Video.'