gpt-oss-120b and gpt-oss-20b are here. Developed with feedback from the open-source community, these models push the frontier of open models at their size.
Hugging Face, NVIDIA, Ollama, vLLM, LM Studio and OpenAI have come together to host this six week hackathon. Your challenge is to apply these models in creative, unexpected ways: build a robot with reasoning, craft a local agent that thrives offline, fine-tune a model into something indispensable, or push the edge with the weirdest hardware you can imagine. Maybe your project will benefit humanity, or maybe it’ll be so unexpected it surprises everyone in the room. However you build, this is your moment to redefine what open models can do.
Get Started
-
Both gpt-oss models are free to download on Hugging Face.
-
Learn more about the models: openai.com/open-models
-
Github: github.com/openai/gpt-oss
-
Check out this microsite of gpt-oss in action: gpt-oss.com
Requirements
What to Build
Apply the gpt-oss models in the most interesting way. Surprise us with something we haven’t seen before! Orient your application into the following categories—we’ll pick one winner for each:
-
Best Overall — The absolutely most interesting application of gpt-oss.
-
Best In Robotics — The most effective use of gpt-oss models in an IRL robot. (Record a video of it in action!)
-
Weirdest Hardware — The most creative and interesting hardware product that uses gpt-oss models.
-
Best Local Agent — The most useful agentic application of gpt-oss with no internet access.
-
Most Useful Fine-Tune — The most effective use of gpt-oss for specialized domains, through fine-tuning.
-
Wildcard — The most unexpected use of gpt-oss that OpenAI didn’t think of.
-
For Humanity — The best application of gpt-oss that benefits all of humanity.
What to Submit
-
Choose the category or categories that best aligns with your project and provide a brief explanation for your choice.
-
Text description that explains the features and functionality of your Project.
-
<3-minute Demonstration video of your Project
-
Provide a URL to your public code repository for judging and testing, with clear indication of gpt-oss model use (e.g., README with testing instructions, sample data, or a link to model weights / training data if you fine-tuned). An OSS/Apache License is encouraged, but not required.
Prizes
Best Overall | Sponsored by Hugging Face
Best in Robotics | Sponsored by NVIDIA
NVIDIA RTX6000 PRO GPU
Weirdest Hardware | Sponsored by Ollama & NVIDIA
$5,000 USD sponsored by Ollama
RTX5090 GPU sponsored by NVIDIA
Best Local Agent | Sponsored by NVIDIA
NVIDIA RTX5090 GPU
Most Useful Fine-Tune | Sponsored by LM Studio
Wildcard—best unexpected use | Sponsored by vLLM
For Humanity | Sponsored by OpenAI
Devpost Achievements
Submitting to this hackathon could earn you:
Judges
Dmitry Pimenov
OpenAI
Casey Dvorak
OpenAI
Dominik Kundel
OpenAI
Vaibhav Srivastav
Hugging Face
Parth Sareen
Ollama
Michael Chiang
Ollama
Mitesh Patel
NVIDIA
Tanay Varshney
NVIDIA
Jay Rodge
NVIDIA
Annie Surla
NVIDIA
Ekaterina Sirazitdinova
NVIDIA
Chris Alexiuk
NVIDIA
Anusha Pant
NVIDIA
Simon Mo
vLLM
Roger Wang
vLLM
Yagil Burowski
LM Studio
Judging Criteria
-
Application of gpt-oss
Does the project apply the open models in an effective way, fitting with the categories? Can other models do the same thing, or does the project showcase the strengths of the models uniquely? -
Design
Is the user experience and design of the project well thought-out, including safety of the user? Is there a balanced blend of frontend and backend? -
Potential Impact
How big of an impact could the project have for its intended audience? How big of an impact could it have beyond the target community and the rest of the world? -
Novelty of the Idea
How creative and unique is the project? Does the concept exist already? If so, how much does the project improve on it?
Questions? Email the hackathon manager
Tell your friends
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
