gemma-4-31B-it-GGUF

gemma-4-31B-it-GGUF

Runs Google's Gemma 4 31B instruction-tuned model locally from Unsloth's GGUF files, with text and image input and a long context. Suits developers who want a strong local reasoning and coding model.

Made by unsloth on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Unsloth's GGUF packaging of google/gemma-4-31B-it, the 31B instruction-tuned member of Google DeepMind's Gemma 4 family. Gemma 4 takes text and image input and writes text. The card describes configurable thinking modes, native function calling and a system role.

Who it suits: Developers and tinkerers who want a large open model on their own machine for reasoning, coding and image questions, and who are comfortable with GGUF tools.

Industry | Who uses it | How they use it | Example
Software | Developers | Run a local coding and reasoning assistant | Ask it to debug a function
Agents | Agent builders | Use native function calling in local agents | Let a local agent call tools
Research | Researchers | Test a family model through a quantized copy | Compare thinking modes on the same prompt

If you only have a small GPU, you can skip it and pick one of the smaller Gemma 4 sizes.

Where to use it:

• llama.cpp and other tools that read GGUF
• Unsloth Studio, which the card says can run and fine-tune Gemma 4
• Windows, macOS or Linux, for downloading the files

Use case | Who | What they get back
Coding help | Developers | Code, fixes and explanations
Image questions | Anyone | Text answers about a picture
Long-context tasks | Analysts | Answers across a context of up to 256K tokens

How to set it up:

• Python with pip, to download the files
• A GGUF runner such as llama.cpp or Unsloth Studio
• Memory for about 15.3 GB at 4-bit by Hugging Face's estimate

1. Install the Hugging Face hub package.
2. Download the GGUF files.
3. Load them in your runner. The card says to re-download after Google's chat template and llama.cpp fixes.

pip install -U huggingface_hub
hf download unsloth/gemma-4-31B-it-GGUF

How to use it: Load the GGUF file in your runner and chat with it. The card points to Unsloth's own guide for running Gemma 4 and notes that multi-token prediction support has been added. Turn thinking on for harder problems.

Example: You load the model and ask it to explain why a function returns the wrong value. It reasons through the code and replies with the cause and a fix.

For newcomers and experts: If you are new, try Unsloth Studio or llama.cpp with one short question. Start without images, then add them.

Gemma 4 comes in dense and mixture-of-experts forms in four sizes: E2B, E4B, 26B A4B and 31B. The medium models offer a 256K context window and the small ones 128K. This copy is a quantized build of the 31B model, so quality can differ slightly from the original.

Download the GGUF files
Load them in llama.cpp or Unsloth Studio
Send text, or images too
The model replies, with optional thinking

Good to know:

• The card says to re-download for the latest chat template and llama.cpp fixes
• Gemma 4 supports over 140 languages
• License: Apache 2.0, with Google DeepMind as the model's author
• Cost to run: none named, you supply the hardware
• Data it reaches: your prompts, images and the files downloaded from Hugging Face
• Not for: video and audio, which the card lists only for the small E2B and E4B models

On Hugging Face, 13 November 2026

• Task: image text to text
• Parameters: 30.7B
• Weights: about 15.3 GB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 428,266
• Likes: 677
• License: apache-2.0
• Based on: google/gemma-4-31B-it
• Quantized versions: 2
• Fine-tunes: 1
• Tags: gguf, gemma4, unsloth, gemma, google, image-text-to-text, imatrix, conversational

Links

• Model: https://huggingface.co/unsloth/gemma-4-31B-it-GGUF
• Files: https://huggingface.co/unsloth/gemma-4-31B-it-GGUF/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download unsloth/gemma-4-31B-it-GGUF

Then follow the usage in its model card. Nothing in this kit runs by itself.