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

Libraries

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

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

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

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ArliAI/gpt-oss-20b-Derestricted")
model = AutoModelForCausalLM.from_pretrained("ArliAI/gpt-oss-20b-Derestricted")
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

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

Install from pip and serve model

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

Use Docker

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

SGLang

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

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

wtf man

by ItzPingCat - opened Dec 1, 2025

Discussion

ItzPingCat

Dec 1, 2025

seems very smart at most things...shits itself at function calling and hallucinates tools and tool outputs

ItzPingCat

Dec 1, 2025

thinking trace looks fine...output not good

base gpt-oss gets the correct answer, which is that web search is not available

albertco2

Dec 1, 2025

seems very smart at most things...shits itself at function calling and hallucinates tools and tool outputs

Same thing happens in Talemate, which is an agentic RP platform. After a lot of tweaking I got it down to 1 in 4 tool / agent calls going fully off the rails. Very unreliable for any agentic system / tool calling. In my experience it's true for all version of GPT-OSS I've tried, except for the base version. Maybe there is a magic set of presets out there for this model, but I haven't found them yet.

ItzPingCat

Dec 1, 2025

But why doesn’t it work

albertco2

Dec 1, 2025

I'm no expert at this, but from my understanding it has something to do with the fine-tuning process. The process of ablation/desrestriction is, right now, not like a tiny surgical incision removing or tweaking alignment. It's more akin to a lobotomy, unrestricted but also less intelligent, or in this case less skilful. Depending on someone's use case they might never really notice it.

ItzPingCat

Dec 2, 2025

Huh

ItzPingCat

Dec 2, 2025

@grimjim thoughts?

grimjim

Dec 3, 2025

From a theoretical standpoint:

With some models, preserving norms/magnitudes is a good thing; others not so much. It's probably a good idea for MoE, but the devil's in the details.

Projecting out the refusal contribution along the harmless direction and orthogonalizing it reduces damage, by minimizing change along the harmless direction itself and around it. That's how that (sub)technique maintains more baseline performance.

This model is heavily quantized given its floating point represenations, so it's possible that requanting post-abliteration could induce additional quant damage, even if intermediate calculations are performed with 32-bit floating point.

ItzPingCat

Dec 4, 2025

Isn't GPT-OSS natively MXFP4? so if you take the mxfp4, up to fp32, perform the ablation and then downscale back to mxfp4, why should there me a major loss in quality