RunPod Storage Upgrade for Gemma 4 | Generated by AI

Home 2026.07

Question: How to upgrade RunPod storage quota so Gemma 4 26B A4B AWQ can fit?

Answer:

Your /workspace is likely a RunPod Network Volume with a small size (around 10GB). You need to increase the Network Volume size, not the pod GPU/container disk. RunPod supports increasing a network volume size later, but you cannot decrease it afterward. (Runpod Documentation)

Option A — RunPod Web UI (recommended)

  1. Open RunPod dashboard:

    • Go to Storage
    • Find your Network Volume

  2. Edit the volume:

    • Increase size from:

      10 GB

      to something like:

      50 GB

  3. Restart/reconnect your Pod.

RunPod’s docs say network volume size can be increased later, but not reduced. (Runpod Documentation)

For your case I would choose:

Gemma 4 26B A4B AWQ
~20 GB model

+ tokenizer
+ SGLang cache
+ LibreChat
+ logs

minimum:
40 GB

comfortable:
50 GB

Option B — API

RunPod exposes an update endpoint:

curl --request POST \
  --url https://rest.runpod.io/v1/networkvolumes/<NETWORK_VOLUME_ID>/update \
  --header "Authorization: Bearer <RUNPOD_API_KEY>" \
  --header "Content-Type: application/json" \
  --data '
{
  "size": 50
}
'

The new size must be larger than the current size. (Runpod Documentation)

Check current volume

Inside pod:

df -h /workspace

You probably see something like:

Filesystem      Size Used Avail
moosefs         10G  10G     0

After resize:

Filesystem      Size Used Avail
moosefs          50G  10G    40G

Important: don’t confuse these three storages

RunPod has:

Container Disk
    |
    | temporary
    | deleted when pod removed


Volume Disk
    |
    | pod-local


Network Volume (/workspace)
    |
    | persistent
    | model storage

For LLM serving, you want Network Volume. (Runpod Documentation)

My recommendation for your setup

Since you are building:

SGLang
   |
   |
LibreChat
   |
   |
local LLM API

I would upgrade once:

10GB  ❌
50GB  ✅
100GB ⭐

50GB is enough for:

Gemma4-26B-A4B-AWQ     20GB
Qwen2.5-14B-AWQ         8GB
embeddings              1-2GB
LibreChat               2GB
cache                   5GB

Then you can swap models without fighting storage every time. (Runpod Documentation)

References:

Back Donate