t5-v1_1-xxl-encoder-gguf

t5-v1_1-xxl-encoder-gguf

A GGUF conversion of Google's T5 v1.1 XXL text encoder, for turning text into embeddings in llama.cpp or feeding image generation models through ComfyUI-GGUF. For developers and image-generation users.

Made by city96 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 conversion of the encoder half of Google's T5 v1.1 XXL model, with about 4.8B parameters. It does not write text. It turns text into embeddings that other tools use. Image generation models often rely on a T5 encoder like this one to read the prompt.

Who it suits: It suits people who run image generation workflows and want a smaller T5 encoder file, and developers who compute text embeddings with llama.cpp. It is a building block, not an app.

Title: Who uses it
Industry | Who uses it | How they use it | Example
Image generation | Workflow builders | Load the encoder through the ComfyUI-GGUF custom node | Pair it with an image model to read prompts
Software | Developers | Compute text embeddings with llama.cpp | Run the llama-embedding tool on a sentence
Hobby use | Owners of modest hardware | Pick a smaller quantized file | Use a lower quant when memory is tight

People who want a chat model can skip it, since it only encodes text.

Where to use it:

• llama.cpp, through its llama-embedding tool
• ComfyUI, through the ComfyUI-GGUF custom node
• Windows, macOS or Linux with Python, to download the files

Title: Top jobs
Use case | Who | What they get back
Prompt encoding | Image-generation users | Text embeddings that an image model can use
Text embeddings | Developers | Vectors for a piece of text
Low-memory setups | Owners of modest hardware | A smaller encoder in a lower quant

How to set it up:

• Python on your computer, for the download
• The huggingface_hub package
• llama.cpp or ComfyUI with the ComfyUI-GGUF node
• About 2.4 GB of disk for a 4-bit file, by estimate

1. Install the Hugging Face hub package.
2. Download the files to your computer.
3. Pick a quant file. The card recommends Q5_K_M or larger.
4. Load it in llama.cpp or the ComfyUI-GGUF node.

pip install -U huggingface_hub
hf download city96/t5-v1_1-xxl-encoder-gguf

How to use it: The card names two ways to use the weights. One is the llama-embedding tool that ships with llama.cpp. The other is the ComfyUI-GGUF custom node, which feeds an image model. The card gives no commands beyond the download.

Example: You download the Q5_K_M file and load it in the ComfyUI-GGUF node next to an image model. When you type a prompt, the encoder turns it into embeddings that the image model uses. The card shows no screenshot or output, so test it in your own workflow.

For newcomers and experts: If you are new, think of this as the part of an image model that reads your prompt. The first thing to try is the Q5_K_M file in an existing workflow, replacing the larger original encoder.

For experts, this is a quantization without an importance matrix, because llama.cpp cannot create one for T5 models. That is why the card advises Q5_K_M or larger, though smaller files can still give decent results when resources are tight.

Download a quant file
Load it in llama.cpp or ComfyUI-GGUF
Text goes into the encoder
Embeddings come out

Good to know:

• Quality: the card recommends Q5_K_M or larger. Smaller files may still work in tight setups.
• No imatrix: the quants are plain, since llama.cpp could not make an importance matrix for T5.
• License: Apache 2.0.
• Cost to run: none named.
• Data it reaches: it downloads files from Hugging Face, and the source names nothing else.
• Not for: generating text, since it is an encoder only.

On Hugging Face, 13 November 2026

• Library: gguf
• Parameters: 4.8B
• Weights: about 2.4 GB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 220,826
• Likes: 552
• License: apache-2.0
• Languages: en
• Based on: google/t5-v1_1-xxl
• Tags: gguf

Links

• Model: https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf
• Files: https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download city96/t5-v1_1-xxl-encoder-gguf

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