embeddinggemma-2 GGUF

embeddinggemma-2 GGUF

Runs a GGUF copy of Google's embeddinggemma-2, a 271M feature extraction model, with the llama command or the Llama app. For developers who want a small text embedding model locally.

Made by ggml-org 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: This is a GGUF copy of google/embeddinggemma-2, a 271M feature extraction model, made by ggml-org. By its name it is a text embedding model. A tool converted it automatically, so it runs in the llama tools.

Who it suits: It suits developers who want a small embedding model on their own machine, for search or similarity features. The card gives few other details.

Title: Who runs it
Industry | Who uses it | How they use it | Example
Software | Developers | Run an embedding model locally with the llama command | A local search feature that needs text vectors
Research | Retrieval researchers | Try a small GGUF embedding model | Compare it with other local embedders
IT | Teams that keep data in-house | Host embeddings on their own hardware | An internal document search

If you need a chat model that writes text, you can skip it.

Where to use it:

• The Llama app
• The llama command in a terminal on Windows, macOS or Linux
• Python on the same machines, for downloading the files by hand

Title: Jobs it does
Use case | Who | What they get back
Local embeddings | Developers | A model served on your machine
Retrieval experiments | Researchers | A small model to test search ideas with

How to set it up:

• The Llama app or the llama command
• Python, only if you download the files by hand

1. Install the Llama app or the llama command.
2. Run the serve command below. It fetches the model by its Hugging Face name.
3. Optionally download the files by hand with the Hugging Face tools.

llama serve -hf ggml-org/embeddinggemma-2-GGUF

pip install -U huggingface_hub
hf download ggml-org/embeddinggemma-2-GGUF

How to use it: Run the serve command in a terminal, or open the model in the Llama app. The card shows nothing beyond that, so check the llama tools for how to send text to a running server.

llama serve -hf ggml-org/embeddinggemma-2-GGUF

Example: You run the serve command in a terminal. The tool fetches the model from Hugging Face and starts it on your machine. The card shows no sample request or output, so what comes back depends on how you call the server.

For newcomers and experts: If you are new, an embedding is a list of numbers that stands for the meaning of a text. The first small thing to try is the serve command above.

If you know the field, this is the original embeddinggemma-2 weights converted to GGUF, at 271M parameters, which is about 136 MB at 4-bit by Hugging Face's estimate. The card gives no dimension count or benchmark, so look at the source model for those.

Run llama serve with the model name
Files come from Hugging Face
The model starts on your machine
Send text and read the vectors

Good to know:

• Conversion: an automatic tool from ggml-org made this copy from the source model.
• Source: the base model is google/embeddinggemma-2.
• Detail: the card has no scores, sizes of the vectors or usage notes.
• License: apache-2.0.
• Cost to run: no API key, account or paid service is named.
• Data it reaches: you download the model from Hugging Face, and the card names no other access.

On Hugging Face, 13 November 2026

• Task: feature extraction
• Parameters: 271M
• Weights: about 136 MB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 125,632
• Likes: 60
• License: apache-2.0
• Based on: google/embeddinggemma-2
• Fine-tunes: 1
• Tags: gguf, quantized, embedding, feature-extraction

Links

• Model: https://huggingface.co/ggml-org/embeddinggemma-2-GGUF
• Files: https://huggingface.co/ggml-org/embeddinggemma-2-GGUF/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download ggml-org/embeddinggemma-2-GGUF

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