model2vec is a Rust crate designed for efficient inference with static embedding models produced using the Model2Vec technique. This approach distills large sentence transformer models into compact, high-performance static embedding models, reducing both model size and computational requirements for inference. The crate is intended for applications that require fast generation of embeddings from text, making it suitable for developers and systems prioritizing speed and efficiency in embedding workflows.
The crate provides an optimized Rust implementation for generating embeddings, supporting fast inference and batch processing. It allows users to encode multiple sentences at once, with options to configure maximum sequence length and batch size during encoding. Models compatible with model2vec can have weights in f32, f16, or i8 formats, stored in safetensors files. Users can load models either from a local directory or by referencing a model ID from the HuggingFace Hub, specifically from the MinishLab collection of pre-trained Model2Vec models. The output embeddings are returned as two-dimensional arrays, with each row corresponding to a sentence and each column to a dimension of the embedding.
The crate is delivered as a Rust library and can be added as a dependency in Rust projects via Cargo. Its documentation notes that it is fully documented and available for the x86_64-unknown-linux-gnu platform. Performance benchmarks indicate that the Rust implementation achieves higher throughput compared to the Python version of Model2Vec, processing approximately 8,000 samples per second in single-threaded CPU tests.
model2vec is released under the MIT license, allowing for open-source use and modification. The documentation also requests that users cite the original Model2Vec project if the methodology or models are used in research or published work. This tool is positioned within the class of static embedding inference engines, specifically optimized for use with Model2Vec-distilled models in Rust environments.
model2vec sits in PulseGate's Other AI category. It focuses on generating fast, compact static embeddings from sentence transformers for efficient NLP applications in Rust. model2vec is an open-source project aimed at rust developers working with NLP and embeddings. The project is open source (MIT). model2vec is available on the web and the command line, and it can be self-hosted.
Behind model2vec is H2CO3, based in Serbia, and the product first shipped in 2025. Across PulseGate's embedding index, model2vec has few near neighbours, marking it as relatively distinct. Among its 5 catalogued features are static embeddings, fast inference, and rust integration.
Latest indexed changes and source events
Other apps tracked under the same category.
