Instructions to use StandardThinking/StandardOne-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use StandardThinking/StandardOne-3B with Transformers:

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

pipe = pipeline("text-generation", model="StandardThinking/StandardOne-3B")
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("StandardThinking/StandardOne-3B")
model = AutoModelForMultimodalLM.from_pretrained("StandardThinking/StandardOne-3B", 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 StandardThinking/StandardOne-3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "StandardThinking/StandardOne-3B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "StandardThinking/StandardOne-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/StandardThinking/StandardOne-3B

SGLang

How to use StandardThinking/StandardOne-3B 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 "StandardThinking/StandardOne-3B" \
    --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": "StandardThinking/StandardOne-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "StandardThinking/StandardOne-3B" \
        --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": "StandardThinking/StandardOne-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use StandardThinking/StandardOne-3B with Docker Model Runner:
```
docker model run hf.co/StandardThinking/StandardOne-3B
```

StandardOne-3B / docs /public-classification-suites.md

MyeongHoJeong

Update model card

815f789 verified 15 days ago

preview code

Raw History Blame Contribute Delete

3.34 kB

Public classification and decision suites

Note. The Standard One 3B v2 figures for these suites are on its model card; the tables below are still those of v1.1.

Measured 25 September 2026 through the served endpoints (merged BF16, SGLang 0.5.20, jev-adapter, native wording, no system prompt, single option order). Each suite: 400 cases sampled with seed 13 from the public test split, one request per case; ECE (10-bin, top label) uses the served temperature (8B T=1.65, 3B T=1.55). Accuracy and macro-F1 are argmax metrics and do not depend on the temperature. banking77 (77 labels) is not runnable: the endpoint accepts at most 26 options per question.

Public datasets

Suite	n	Acc 8B	Acc 3B	Macro-F1 8B	Macro-F1 3B	ECE 8B	ECE 3B
AG News (4 topics)	400	84.2	83.0	82.9	81.4	0.084	0.094
DAIR Emotion (6 labels)	400	56.5	58.8	46.6	48.6	0.191	0.064
SST-5 (5-level sentiment, ordinal score)	400	—	—	—	—	—	—
Typed decisions (400 cases, 2,000 decisions)	400	69.3	69.0	60.2	58.9	0.056	0.048

Application workflows (400 cases each, public sources)

Suite	n	Acc 8B	Acc 3B	Macro-F1 8B	Macro-F1 3B	ECE 8B	ECE 3B
Email spam	400	92.2	93.0	92.2	93.0	0.047	0.045
Phishing	400	86.0	89.8	84.2	88.6	0.048	0.037
LLM guardrails (jailbreak vs benign)	400	50.7	57.5	50.5	56.7	0.314	0.181
Moderation (toxic vs benign)	400	85.0	81.5	85.0	81.4	0.079	0.046
RAG passage relevance	400	60.2	57.2	56.9	53.7	0.307	0.289
Support triage (10 queues)	400	35.2	39.5	36.1	37.5	0.435	0.359
Model routing (domain)	399	93.0	83.7	56.5	52.8	0.048	0.046

MASSIVE intent, 20 options (chance = 5 %)

Language	n	Acc 8B	Acc 3B	Macro-F1 8B	Macro-F1 3B	ECE 8B	ECE 3B
en	400	88.0	85.5	61.7	54.3	0.024	0.055
ja	400	87.0	86.0	60.1	65.2	0.032	0.037
zh	400	86.0	84.0	63.2	60.7	0.042	0.054
es	400	82.0	82.0	55.5	53.5	0.025	0.064
fr	400	84.5	82.8	59.2	54.2	0.040	0.061
de	400	85.8	83.8	62.6	52.8	0.039	0.043
pt	400	83.5	80.5	53.4	48.5	0.044	0.056
ru	400	85.0	81.8	62.0	55.5	0.038	0.062
ko	400	88.2	83.8	67.2	58.0	0.034	0.029

XNLI, 3 labels

Language	n	Acc 8B	Acc 3B	Macro-F1 8B	Macro-F1 3B	ECE 8B	ECE 3B
en	400	63.5	69.2	64.2	68.7	0.136	0.080
es	400	57.2	64.0	57.5	63.7	0.186	0.057
fr	400	61.0	64.2	61.3	63.7	0.152	0.067
de	400	60.5	60.5	60.3	60.2	0.160	0.024
ru	400	57.2	60.8	56.3	60.2	0.212	0.047
zh	400	61.0	53.2	61.0	52.6	0.169	0.061

Option-order robustness

Share of cases whose answer changes when the options are presented in a permuted order (lower is better).

Suite	8B	3B
AG News (4 topics)	0.033	0.037
DAIR Emotion (6 labels)	0.072	0.185
massive_intent.en	0.133	0.135

SST-5 score MAE on the 0–4 scale: 8B 0.566, 3B 0.653.