Instructions to use StandardThinking/StandardOne-3B-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use StandardThinking/StandardOne-3B-FP8 with Transformers:

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

pipe = pipeline("text-generation", model="StandardThinking/StandardOne-3B-FP8")
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-FP8")
model = AutoModelForMultimodalLM.from_pretrained("StandardThinking/StandardOne-3B-FP8", 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-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "StandardThinking/StandardOne-3B-FP8"
# 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-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/StandardThinking/StandardOne-3B-FP8

SGLang

How to use StandardThinking/StandardOne-3B-FP8 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-FP8" \
    --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-FP8",
		"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-FP8" \
        --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-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use StandardThinking/StandardOne-3B-FP8 with Docker Model Runner:
```
docker model run hf.co/StandardThinking/StandardOne-3B-FP8
```

MyeongHoJeong commited on Sep 27

Commit

093f76c

verified ·

1 Parent(s): 860e6f5

Restore FP8 model card wording

Browse files

Files changed (1) hide show

README.md +5 -7

README.md CHANGED Viewed

@@ -16,10 +16,8 @@ tags:
 # StandardOne-3B-FP8
-> **Version scope (2026-09-27).** This FP8 checkpoint was quantized from [StandardOne-3B BF16 revision 68dafd17](https://huggingface.co/StandardThinking/StandardOne-3B/tree/68dafd17ead9b8cf6f85c4f08f7f2f3a1e7b9e5c) on 2026-09-25, before the BF16 v2 update on 2026-09-26. The validation and temperature below apply to this earlier checkpoint; this file is not FP8 v2. Its training provenance follows that pinned source revision, not the current BF16 v2 card.
 StandardOne-3B-FP8 is an FP8 (compressed-tensors, float8_e4m3 weights, dynamic per-token activations)
-quantization of the pinned StandardOne-3B BF16 decision model. The language-model linear
 projections (q/k/v/o, gate/up/down) are quantized per-channel FP8 E4M3 with dynamic FP8
 activations (llm-compressor's data-free `FP8_DYNAMIC` recipe, no calibration data required);
 the vision tower, multi-modal projector, embeddings and lm_head are left unquantized in BF16.
@@ -30,9 +28,9 @@ compressed-tensors 0.19.0); results below.
 ## Validation
 Served through SGLang 0.5.20 and `jev-adapter` (native wording, no system prompt, one option order) on the
-same items as the pinned BF16 source, measured 2026-09-25. Accuracy is argmax and does not depend on temperature.
-| Suite | BF16 (pinned source) | FP8 (this repository) |
 |---|---:|---:|
 | JevBench public easy (48) | 100.00 % | 100.00 % |
 | JevBench public standard (72) | 88.89 % | 88.89 % |
@@ -42,5 +40,5 @@ same items as the pinned BF16 source, measured 2026-09-25. Accuracy is argmax an
 | stated-distribution probability (1,036) | 80.50 % | 79.05 % |
 | hard proxy (600) | 42.17 % | 41.67 % |
-Temperature refit on this checkpoint's own served probabilities (same held-out calibration data as the pinned BF16
-source): T = 1.3 (pinned BF16 source: T = 1.55). Serve with `--default-temperature 1.3`.

 # StandardOne-3B-FP8
 StandardOne-3B-FP8 is an FP8 (compressed-tensors, float8_e4m3 weights, dynamic per-token activations)
+quantization of the released StandardOne-3B decision model. The language-model linear
 projections (q/k/v/o, gate/up/down) are quantized per-channel FP8 E4M3 with dynamic FP8
 activations (llm-compressor's data-free `FP8_DYNAMIC` recipe, no calibration data required);
 the vision tower, multi-modal projector, embeddings and lm_head are left unquantized in BF16.
 ## Validation
 Served through SGLang 0.5.20 and `jev-adapter` (native wording, no system prompt, one option order) on the
+same items as the BF16 release, measured 2026-09-25. Accuracy is argmax and does not depend on temperature.
+| Suite | BF16 (StandardOne-3B) | FP8 (this repository) |
 |---|---:|---:|
 | JevBench public easy (48) | 100.00 % | 100.00 % |
 | JevBench public standard (72) | 88.89 % | 88.89 % |
 | stated-distribution probability (1,036) | 80.50 % | 79.05 % |
 | hard proxy (600) | 42.17 % | 41.67 % |
+Temperature refit on this checkpoint's own served probabilities (same held-out calibration data as the BF16
+release): T = 1.3 (BF16 release: T = 1.55). Serve with `--default-temperature 1.3`.