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

Libraries

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

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

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

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("PeetPedro/gpt-oss-120b-heretic")
model = AutoModelForCausalLM.from_pretrained("PeetPedro/gpt-oss-120b-heretic", 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]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

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

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "PeetPedro/gpt-oss-120b-heretic"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PeetPedro/gpt-oss-120b-heretic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

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

SGLang

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

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

gpt-oss-120b-heretic

Stage abliteration output of an open abliterate → SFT → RFT → RLVR coding-model pipeline. Refusal directions removed via Heretic weight surgery (no gradients).

Base model: unsloth/gpt-oss-120b-BF16
Chat format: harmony (gpt-oss)
Pipeline: https://github.com/peterlodri-sec/heretic-coder-pipeline

Evaluation

Metric	Value
`refusal_rate`	0.0267
`kl_divergence`	0.1231
`mmlu_delta`	(not measured)
`gsm8k_delta`	(not measured)

Intended use & responsible use

This model is abliterated (uncensored). Refusal behaviour has been deliberately reduced, so it will not reliably decline unsafe or disallowed requests. It is intended for internal, gated engineering use behind verify-before-merge and your own moderation / authorization layer — never user-facing without an independent safety layer. Weights are private / gated. You own the outputs; use it lawfully.

Provenance

Built with Heretic (abliteration), Unsloth + TRL (SFT/RFT/RLVR). See the pipeline repo for the exact stage configs, gates, and the reproducible harness.

No benchmark numbers are claimed beyond the table above; evaluate on your own tasks.

Downloads last month: 28

Safetensors

Model size

117B params

Tensor type

BF16

Model tree for PeetPedro/gpt-oss-120b-heretic

Base model

openai/gpt-oss-120b

Finetuned

unsloth/gpt-oss-120b-BF16

Finetuned

(1)

this model