---
title: "What does GPU VRAM Allocated mean?"
description: "How ARX estimates each deployed model's VRAM usage, why a deployment is blocked when a card runs short, and where to check the remaining capacity"
canonical_url: "https://docs.ardge.com/en/knowledge/arx-os/gpu-vram-allocated"
---
# What does GPU VRAM Allocated mean?

> How ARX estimates each deployed model's VRAM usage, why a deployment is blocked when a card runs short, and where to check the remaining capacity

## Short answer

GPU VRAM allocated is the estimated VRAM usage ARX calculates for every deployed model. When several models run on the same GPU card, the system uses this figure to track how the card's capacity is committed, which reduces the chance that models compete for memory, run out, and take the service down.

## Details

One GPU card can run several models at once. If the VRAM they actually consume adds up to more than the card holds, the card runs out of memory and the service goes down. ARX estimates each model's VRAM requirement at deployment time and records it as that model's GPU VRAM allocated. The system therefore knows how much of each card is committed and how much is left before anything is over-committed.

### What happens when a card doesn't have enough left?

The system blocks the deployment outright. If a GPU card's remaining capacity can't hold the new model's estimated usage, that model can't be deployed to that card.

So the remaining capacity on the card tells you directly what you can do: when it exceeds the target model's estimated requirement, you can deploy the model there. When it doesn't, use a different GPU card, or stop a model on that card that you no longer need to free up its allocation.

## How to check

Go to **AI Model Center › Dashboard › GPU**. Each GPU card shows the estimated usage of the models on it, along with the card's current remaining capacity.

Before deploying a new model, confirm that the target GPU card has enough capacity left for the VRAM it needs. To see how much VRAM a model needs, open **Model Hub** and click the model card to expand its details.


## Sitemap

See the full [sitemap](https://docs.ardge.com/sitemap.md) for all pages.