To the Hugging Face Team and Board,
I am writing to express the community’s extreme frustration regarding the recent layout overhauls to the Hugging Face free tier. Specifically, the decisions to completely eliminate the free CPU Basic instance flavor, lock down both the Docker and Gradio SDKs for unpaid personal/organization spaces, and force everyone onto a highly restricted ZeroGPU tier.
The current ZeroGPU restrictions—limiting free tier users to a handful of requests or a few minutes of compute per day—make it completely impossible to debug code, test multi-model routing architectures, or run sustained experiments. Frankly, because of these strict, unusable limits, the community does not care about the ZeroGPU allocation. We would gladly trade it away to have the platform work the way it was originally supposed to.
Hugging Face built its massive reputation by being the ultimate hub for accessible open-source AI. Locking out independent developers and free organizations from creating standard CPU-driven spaces completely kills grassroots innovation.
We strongly urge the product team and board to:
Revert the hardware changes and make the standalone CPU Basic tier completely free and unmetered again.
Remove the paywalls on SDK configurations for both personal accounts and unpaid organizations, allowing raw Docker and Gradio layouts to deploy on basic CPU hardware without restrictions.
Accessibility is what made this platform great. Please stop locking out the very community that helped build Hugging Face.
Sincerely, An Independent Open-Source Developer