Kaggle and Hugging Face users are part of one AI community. That’s why we look forward to announcing plans to bring platforms and communities closer to AI developers, to deliver services everywhere.
Starting today, Kaggle is launching an integration that increases visibility and discoverability for directly hugging face models with Kaggle.
How to get started
You can navigate to Kaggle from hugging the face model and vice versa. Start by visiting the face model page for hugs like QWEN/QWEN3-1.7B. To use it with a Kaggle Notebook, you can click “Use this model” and select “Kaggle” to open a Kaggle notebook with pre-entered code snippets and load the model. You can do the same from Kaggle’s embraced face model page by clicking the “Code” button.
When you run a notebook in Kaggle that references a model that hugs Face Hub, it automatically generates a face model page to hug if it doesn’t already exist. There is no need to make any special changes to the code. Additionally, when you publish a notebook, it will automatically appear in the Code tab of the Kaggle model page.
Discover how to hug a facial model and explore examples of all community of public notes at one location in Kaggle at https://www.kaggle.com/models. As more embraced face models are used in Kaggle, there will be more examples of code related to the number of models that can be explored for inspiration.
When hugging face models in Kaggle, it’s easy to embrace your face to explore additional details, metadata, and community use to hug face spaces, discussions and more. Simply click “Open with Face” on the Kaggle model page.
How does this work with private and consented embracing face models?
If you use a private hug face model in your Kaggle Notebook, authenticate through a face account that hugs as usual (please add hf_token to the “Add-ons>secrets” menu in the notebook editor). Kaggle does not generate a face model page to hug.
If you want to access the consent gate model on your Kaggle Notebook, you must request access using a face account that you want to hug and follow the prompts on your browser’s hugging face model page as normal. Hugging Face has documentation to guide you through this process. Otherwise, integration works just like a non-gate model.
What’s next?
We are actively working on solutions that seamlessly use embracing face models in Kaggle competitions that require offline notebook submissions. This will take several more months, but I believe the wait is worth it.
To understand why getting this part of integration is so important, you can read Kaggle’s position on “AI competition as the gold standard for empirical rigor of genai evaluations”! However, TL;DR-Kaggle is extremely sensitive to data leakage and the effects on model contamination. Our goal is to design this integration to maintain competitive integrity and a key role in the industry, while allowing Kaggle competitors to build seamless access with Face’s best models.
In the meantime we’d love to hear your feedback – share your thoughts and ideas here!
Happy kaggling!
