openai
/

gpt-oss-120b

Text Generation

8-bit precision

Model card Files Files and versions

Instructions to use openai/gpt-oss-120b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use openai/gpt-oss-120b with Transformers:

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

pipe = pipeline("text-generation", model="openai/gpt-oss-120b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("openai/gpt-oss-120b")
model = AutoModelForCausalLM.from_pretrained("openai/gpt-oss-120b", device_map="auto")
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
AMD Developer Cloud
Local Apps Settings

How to use openai/gpt-oss-120b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "openai/gpt-oss-120b"
# 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-120b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/openai/gpt-oss-120b

How to use openai/gpt-oss-120b 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-120b" \
    --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-120b",
		"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-120b" \
        --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-120b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use openai/gpt-oss-120b with Docker Model Runner:
```
docker model run hf.co/openai/gpt-oss-120b
```

Resources

View closed (40)

Update README.md

#119 opened 12 months ago by

new version miss file,can't startup the model gpt-oss

#118 opened 12 months ago by

Fine tune 120b at 8 H100s getting cuda OOM error

#117 opened 12 months ago by

68.4 on Aider Polyglot with reasoning_effort: high

#116 opened 12 months ago by

fix missing the `{% generation %}` keyword while using tokenizer.apply_chat_template(...return_assistant_tokens_mask=True)

#112 opened 12 months ago by

Two clarifications on gpt-oss-120B hardware (fine-tuning vs inference, MoE VRAM)

#110 opened 12 months ago by

gpt-oss is actually good. even on less common benchmark

#109 opened 12 months ago by

groupfairnessllm

Please restart openai/gpt-oss-120b endpoint + confirm logprobs support

#107 opened 12 months ago by

Model output with FP16

#106 opened 12 months ago by

BEST open source 120b level model.. ever

#104 opened 12 months ago by

Multimodal support missing

#101 opened 12 months ago by

LiveCodeBench evaluation

#100 opened 12 months ago by

How to reproduce the results on hle(Humanity's Last Exam)?

#95 opened 12 months ago by

ImportError: /lib64/libc.so.6: version `GLIBC_2.32' not found

#86 opened 12 months ago by

Why is the performance worse on the 20B than on the 120B

#84 opened 12 months ago by

How to turn off thinking mode

#82 opened 12 months ago by

gpt-oss120b does not actually support 131072 output tokens due to openai policies embedded limiting ouput

#81 opened 12 months ago by

Run GPT-OSS-120B with just Single A100 (80GB)

#80 opened 12 months ago by

Please move metal/ and original/ to their own repos

#78 opened 12 months ago by

gp-oss-120b — Exceptional Reasoning, Not Yet AGI Scale

#77 opened 12 months ago by

BertrandCabotPro

Qwen3 beat gpt-oss with just 0.6B with good quality enough to be usable

#75 opened 12 months ago by

Hallucinates System-promp

#74 opened 12 months ago by

Can gpt-oss support local vllm deployment on a100 GPU?

#73 opened 12 months ago by

gpt_oss error when running on kaggle

#72 opened 12 months ago by

is this guid prompt is still valid ?

#71 opened 12 months ago by

[Discussion] gpt-oss-120b hangs indefinitely ("thinking...") when using YaRN rope scaling to extend context length

#70 opened 12 months ago by

Errors in chat template compared to spec

#69 opened 12 months ago by

Possible PEP 660 violation in `_build/gpt_oss_build_backend/backend.py`

#66 opened 12 months ago by

Estimated Thinking/Reasoning Token Usages For Each Modes

#65 opened 12 months ago by

running mxfp4 on H100 using tranformers with triton_kernel: make_default_matmul_mxfp4_w_layout not found

#64 opened 12 months ago by

🚀 Best Practices for Evaluating GPT-OSS Models: Speed & Benchmark Testing Guide

#62 opened 12 months ago by

How do I serve a model in the original folder as bf16 in VLLM?

#60 opened 12 months ago by

Model Performance

#59 opened 12 months ago by

Disgusting, maximally censored model!

#56 opened 12 months ago by

🔥🔥🔥中文测评视频

#55 opened 12 months ago by

Llama, Mistral, Gemma… and now OpenAI enters the hunger games. 🐎⚔️

#54 opened 12 months ago by

Request: FP8 / BF16 version of model?

#53 opened 12 months ago by

кал

#51 opened 12 months ago by

how to disable the reasoning mode?

#50 opened 12 months ago by

Enterprise AI factory OS

#47 opened 12 months ago by

DavidSteinbauer

How to use different reasoning effort in the example?

#45 opened 12 months ago by

Structured Outputs

#44 opened 12 months ago by

Problems with Metal

#42 opened 12 months ago by

[v1 engine][flash_attn backend] TypeError: flash_attn_varlen_func() got an unexpected keyword argument 's_aux' when running gpt-oss-120b on H200

#41 opened 12 months ago by

openai_harmony.HarmonyError: error downloading or loading vocab file: failed to download or load vocab file

#39 opened 12 months ago by

Cooooool !大模型本地部署，能力评测，幻觉评测微信交流群，欢迎感兴趣的朋友加入

#38 opened 12 months ago by

I feel unsafe

#37 opened 12 months ago by

Deploy gpt-oss models in your own AWS account using vLLM and Tensorfuse

#36 opened 12 months ago by

Local Installation Video and Testing - Step by Step

#34 opened 12 months ago by

vLLM FlashAttention3 with A6000

#33 opened 12 months ago by