Instructions to use openai/gpt-oss-20b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use openai/gpt-oss-20b with Transformers:

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

pipe = pipeline("text-generation", model="openai/gpt-oss-20b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("openai/gpt-oss-20b")
model = AutoModelForCausalLM.from_pretrained("openai/gpt-oss-20b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
HuggingChat
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use openai/gpt-oss-20b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "openai/gpt-oss-20b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "openai/gpt-oss-20b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/openai/gpt-oss-20b

SGLang

How to use openai/gpt-oss-20b 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 "openai/gpt-oss-20b" \
    --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": "openai/gpt-oss-20b",
		"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 "openai/gpt-oss-20b" \
        --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": "openai/gpt-oss-20b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use openai/gpt-oss-20b with Docker Model Runner:
```
docker model run hf.co/openai/gpt-oss-20b
```

Does transformers utilize PyTorch SDPA's flash_attention for openai/gpt-oss-20b?

#89

by NooBaymax - opened Aug 8, 2025

Discussion

NooBaymax

Aug 8, 2025

I'm investigating if the flash_attention backend from PyTorch's scaled_dot_product_attention (SDPA) is leveraged when running the openai/gpt-oss-20b model via the transformers library. How can we verify this behavior? I'm looking for methods, code snippets, or official documentation that confirm whether this optimization is active by default or if specific configurations are required to enable it.

reach-vb

OpenAI org Aug 8, 2025

You can switch between various attn_implementations like here: https://huggingface.co/docs/transformers/en/main_classes/model#transformers.PreTrainedModel.from_pretrained.attn_implementation

NooBaymax

Aug 15, 2025

You can switch between various attn_implementations like here: https://huggingface.co/docs/transformers/en/main_classes/model#transformers.PreTrainedModel.from_pretrained.attn_implementation

it looks like gpt-oss model doesn't support sdpa,

i got
raise ValueError(
ValueError: GptOssForCausalLM does not support an attention implementation through torch.nn.functional.scaled_dot_product_attention yet. Please request the support for this architecture: https://github.com/huggingface/transformers/issues/28005. If you believe this error is a bug, please open an issue in Transformers GitHub repository and load your model with the argument attn_implementation="eager" meanwhile. Example: model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="eager")

RonanMcGovern

Sep 20, 2025

•

edited Sep 20, 2025

Edit / Correction:

I am able to load the model with FA2 and it works to reduce memory usage on OSS:

model = AutoModelForCausalLM.from_pretrained(
   model_slug,
   torch_dtype=torch.bfloat16,          # bf16 on H100/H200
   device_map="auto",
   attn_implementation="flash_attention_2",
)

I upgraded to the latest transformers. Note fyi that this appears not (yet?) to work on unsloth - for reasons I don't understand.

Original post:

Yeah @reach-vb this is a pretty big bottleneck meaning you can't really train OSS on reasoning.

SPDA isn't supported for OSS (you can see the fallback to eager logged explicitly when you load with unsloth).

It would be nice if attn_implementation set to FA2 worked, but I have tried that and it also just causes a fallback to eager.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment