Luxical-One

Luxical-One is a small lexical-dense text embedding model distilled from Snowflake/snowflake-arctic-embed-m-v2.0. It is optimized for high-throughput embedding without GPU acceleration, and it is trained for English-language symmetrical full-document text similarity and classification applications.

Read the full story of Luxical in our blog post.

Want to go deeper with Luxical models? Check out the repo.

Quickstart

Luxical-One includes a basic HuggingFace integration that supports model inference. GPU-based inference is not supported (GPU "acceleration" would not provide a meaningful boost on most systems due to how fast and small the model is anyhow).

You must have the luxical package installed to run this model, and you must trust the remote code bundled with the model. Currently only cpython versions 3.11-3.13 and macos/linux operating systems are supported.

To install luxical:

pip install luxical

To load and use the model in transformers:

from transformers import AutoModel

example_text = "Luxical integrates with Huggingface."
luxical_one = AutoModel.from_pretrained("datologyai/luxical-one", trust_remote_code=True)
embeddings = luxical_one([example_text]).embeddings

Or via sentence-transformers:

from sentence_transformers import SentenceTransformer

example_text = "Luxical integrates with Huggingface."
luxical_one = SentenceTransformer("datologyai/luxical-one", trust_remote_code=True)
embeddings = luxical_one.encode(example_text)