Text · google
gemma-4-31B-it
Google's Gemma 4 31B instruct model. A 4-bit estimate is the first number to compare with your card. The full repository is larger.
Freshness uses the repository open date. This record is 235 days old, opened 2026-03-11. Popularity uses 8,456,004 downloads and 4,253 likes listed by Hugging Face. Checked 2026-11-01. Neither number grades answers. Formulas.
Hugging Face reports 175.7 GB of storage for the whole repository. That figure can include several packs, so the fit table does not use it as the size of one file.
Will it fit?
Save a machine on the hardware page and this block fills in. The table already has the sizes.
| Pack | Size | Saved machine |
|---|---|---|
| Weight shards summed | 62.5 GB weights + 4.5 GB allowance. File: 2 shard files in the repository root. | Choose a machine |
| Q4 estimate | 17.2 GB weights + 4.5 GB allowance. | Choose a machine |
| Q8 estimate | 32.8 GB weights + 4.5 GB allowance. | Choose a machine |
| FP16 estimate | 62.5 GB weights + 4.5 GB allowance. | Choose a machine |
Estimate rows use the parameter count. A measured row names a file in the repository. The allowance is fixed: it stands in for context or activations, and it is not a measurement from a run.
How to run it
Vision models take images as well as text, so the allowance on the fit row is larger than a text-only model of the same parameter count. Start from the publisher repository and confirm which file is the checkpoint.
Publisher repository on Hugging Face
Others in Text
Qwen3.8-27B · Qwen
bge-m3 · BAAI
Qwen3-8B · Qwen
This directory is assembled from public model records. Freshness is age. Popularity is downloads and likes. A fit figure is an estimate from a parameter count or from a published file size, not a run on our hardware. How the numbers are made. AI disclosure.