GPT OSS 20B Heretic Chat

GPT OSS 20B Heretic Chat

Lets you talk to the gpt-oss-20b-heretic language model in your browser. It runs on ZeroGPU and shows only the final answer, for people who want to try this community model.

Made by locojefe on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space is a chat demo for a community language model called gpt-oss-20b-heretic. You type a message in your browser and the model answers. The app loads the exact model in BF16 through Transformers. Inference runs on ZeroGPU rather than through an outside provider, and no other model stands in for it.

Who it suits: People who want to try this model quickly, without setting up a GPU of their own.

Industry | Who uses it | How they use it | Example
AI research | People who compare open models | Chat with the model to judge its answers | Sending the same question to several models and comparing replies
Software | Developers choosing a model | Test how the model handles short prompts | Trying a coding question before downloading the weights
Education | Students learning about language models | Watch how a 20B model answers in plain chat | Asking a few general questions and reading the replies

If you need long conversations or long answers, you can skip it, because the demo caps both.

Where to use it:

• A web browser, on the Space's page at Hugging Face
• Python 3.12 on Windows, macOS or Linux, for a copy you run yourself

Use case | Who | What they get back
Try the model in chat | Curious users | A final answer to each message
Check the model before wider use | Developers | A quick sense of how it replies
Run a local copy of the demo | People who know Python | The same Gradio app on their own computer

How to set it up: Online, nothing needs installing. A copy on your own computer needs a few things first.

• Python 3.12
• The pip tool
• The packages listed in requirements.txt

1. Install the Hugging Face hub package.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Install Gradio at the pinned version.
5. Install the packages from requirements.txt.
6. Start app.py.

pip install -U huggingface_hub
hf download locojefe/gpt-oss-20b-heretic-demo --repo-type space --local-dir gpt-oss-20b-heretic-demo
cd gpt-oss-20b-heretic-demo
pip install gradio==6.30.0
pip install -r requirements.txt
python app.py

The source does not say what hardware a local run needs, so check that before you try it.

How to use it: Open the Space's page on Hugging Face and type a message into the chat. The model replies with its final answer only. Reasoning tokens are hidden but still count toward the output budget. Each conversation is capped at 3,072 input tokens and 1,024 new tokens. To run the app yourself, start it with Python.

python app.py

Example: You type "Explain what a hash table is in two sentences." The model replies with a short explanation and nothing else. You do not see its reasoning, and the reply stops once it reaches the 1,024 new-token limit.

For newcomers and experts: This is a chat page for one specific model. To start, open the Space in your browser and send a short question. Your messages stay in your browser session, and the app does not keep a transcript.

Under the hood the app is Gradio 6.30.0 on Python 3.12. It pins one model revision and applies the model's Harmony chat template. Because reasoning tokens use up the same output budget as the answer, long reasoning can leave less room for the reply. Hugging Face queue policies apply on ZeroGPU.

Open the Space in a browser
Send a message in the chat
Model runs on ZeroGPU
Only the final answer is shown

Good to know:

• License: Apache 2.0 for the Space and for the upstream model weights
• Cost to run: no API key or paid service is named, and Hugging Face infrastructure and queue policies apply
• Data it reaches: conversations stay in the browser session and the app does not save transcripts, and it loads the model p-e-w/gpt-oss-20b-heretic
• Not for: long inputs or long answers, and OpenAI did not make this demo

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/locojefe/gpt-oss-20b-heretic-demo
• Files: https://huggingface.co/spaces/locojefe/gpt-oss-20b-heretic-demo/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.