bge-small-en-v1.5-GGUF

bge-small-en-v1.5-GGUF

Turns English text into embeddings for search and retrieval, in a small GGUF file of 33.2M parameters. For developers who build search or retrieval apps on modest hardware.

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: bge-small-en-v1.5-GGUF is a GGUF copy of the small English BGE embedding model, shared by unsloth, with 33.2M parameters. BGE is short for BAAI General Embedding. An embedding model turns text into numbers so that similar passages can be found by comparing them. The card text comes from the FlagEmbedding project.

Who it suits: Developers who build search or retrieval features over English text are the main fit, and the small file suits modest hardware. NLP researchers can use it as a compact embedding baseline.

Industry | Who uses it | How they use it | Example
Search | Software developers | Turn English text into vectors for similarity search | Find notes that relate to a question
AI apps | App builders | Retrieve passages to give a language model as context | Fetch the best passages for a chatbot answer
Research | NLP researchers | Compare small embedding models | Use it as a compact baseline

If your text is not English, or you need longer inputs or other retrieval methods, you can skip it, because the card points to bge-m3 for those.

Where to use it: Use it from Python, or in tools that read GGUF files.

• Python scripts that use the sentence-transformers library
• Windows, macOS or Linux, with Python for the download
• Langchain, which the card says BGE models are integrated into

Use case | Who | What they get back
Semantic search | Developers | An embedding per text, to compare by similarity
Retrieval for LLMs | App builders | The closest passages for a query
Re-ranking pipelines | Search engineers | A first-stage retriever that a reranker can refine

How to set it up: You need Python and only a little disk space.

• Python with pip
• Room for the file, about 17 MB at 4-bit as an estimate

1. Install the Hugging Face hub package.
2. Download the model files.
3. Read the Usage section of the card for the exact calls.

pip install -U huggingface_hub
hf download unsloth/bge-small-en-v1.5-GGUF

How to use it: You give the model English text and take its embedding back. To find related passages, embed your query and your passages, then compare the embeddings for similarity. The card text shown here gives no run command beyond the download.

Tool | What it does
bge-m3 | The card's suggestion when you need other languages, longer inputs or extra retrieval methods
bge-reranker-base | A cross-encoder that re-ranks the top documents an embedding model returns
bge-reranker-large | The larger cross-encoder for the same re-ranking job

Example: You embed the question about resetting a password and a few help-center paragraphs, then compare them. The paragraph whose embedding sits closest to the question's embedding is the best match.

For newcomers and experts: An embedding model turns a sentence into a list of numbers, so that sentences with close meanings get close numbers. A good first try is to download the files and compare two short sentences that mean nearly the same thing.

This is the v1.5 small model of the BGE family, which the card says eases the spread of similarity scores and retrieves better without an instruction. It has 33.2M parameters and is packaged as GGUF by unsloth. The trade-off is scope, since it covers English only and the card sends longer inputs to bge-m3.

English text goes in
The model produces an embedding
Embeddings are compared by similarity
A reranker can refine the top results

Good to know: The model is small and English only.

• Language: English
• Format: GGUF, with about 17 MB of weights at 4-bit as an estimate
• Scope: for other languages, longer inputs or extra retrieval methods, the card points to bge-m3
• License: MIT
• Cost to run: none named
• Data it reaches: the English text you embed. The card names no account or web service
• Not for: other languages, longer inputs or other retrieval methods, per the card

On Hugging Face, 13 November 2026

• Task: feature extraction
• Library: sentence-transformers
• Parameters: 33.2M
• Weights: about 17 MB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 165,239
• Likes: 10
• License: mit
• Languages: en
• Tags: sentence-transformers, gguf, feature-extraction, sentence-similarity, transformers, mteb, model-index

Links

• Model: https://huggingface.co/unsloth/bge-small-en-v1.5-GGUF
• Files: https://huggingface.co/unsloth/bge-small-en-v1.5-GGUF/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download unsloth/bge-small-en-v1.5-GGUF

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