# Zerogpu disk space limits? **URL:** \ **Category:** Spaces\ **Created:** [February 2, 2026, 4:50pm UTC](https://discuss.huggingface.co/t/zerogpu-disk-space-limits/173028 "2026-02-02T16:50:34Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![ejschwartz](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/ejschwartz/32/16902_2.png) [@ejschwartz](https://discuss.huggingface.co/u/ejschwartz)\ **Post date:** [February 2, 2026, 4:50pm UTC](https://discuss.huggingface.co/t/zerogpu-disk-space-limits/173028/1 "2026-02-02T16:50:34Z")
I am trying to create a [xlarge](https://huggingface.co/docs/hub/en/spaces-zerogpu#gpu-size-selection) zerogpu space for a fairly large (22B / ~80 GiB) model. Unfortunately, I seem to be running out of disk space during the tensor packing step. I attempted to purchase more storage, only to find that this is [not possible anymore for new spaces.](https://discuss.huggingface.co/t/spaces-persistent-storage-upgrade-not-accessible/171226/11) Is there any way to load such a large model using zerogpu now? If not, is it possible with paid spaces? I actually couldn’t find any documentation stating whether paid spaces provide more ephemeral disk space. @meganariley ---
**Author:** ![John6666](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/john6666/32/27664_2.png) [@John6666](https://discuss.huggingface.co/u/John6666)\ **Post date:** [February 2, 2026, 11:47pm UTC](https://discuss.huggingface.co/t/zerogpu-disk-space-limits/173028/2 "2026-02-02T23:47:30Z")
It seems like GPU space can sometimes offer quite a large amount of disk capacity?: [Using GPU Spaces](https://huggingface.co/docs/hub/spaces-gpus#gpu) But yeah it’s not for Zero GPU Space… ---
**Author:** ![ejschwartz](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/ejschwartz/32/16902_2.png) [@ejschwartz](https://discuss.huggingface.co/u/ejschwartz)\ **Post date:** [February 3, 2026, 2:20pm UTC](https://discuss.huggingface.co/t/zerogpu-disk-space-limits/173028/3 "2026-02-03T14:20:05Z")
@John6666 Thanks, I looked at that table and somehow did not see “disk”. It’s good to know that is an option at least. But still, this definitely lowers the utility of zerogpu if there is no way around it. I’m working on a quantization to see if that will fit…