MinishLab/model2vec

Health v2 · maintenance only

Adoption not part of the score

2186 stars · 124 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Model2Vec is a Python library that distills any sentence transformer into a tiny, fast static embedding model by computing one fixed vector per token and averaging them at inference time. It reduces model size up to 50x and speeds up inference up to 500x on CPU with only a small drop in quality.

Use cases

• generate text embeddings fast on cpu
• distill a sentence transformer into a small static model
• build a lightweight retrieval or rag system
• classify documents at high throughput without gpu
• deduplicate or cluster large text datasets
• embed large corpora like wikipedia quickly
• reduce embedding model size for edge deployment

When to choose

• you need very fast, cheap embeddings on cpu
• you want small models for retrieval, classification, or clustering
• you need to deploy embeddings without gpu or heavy dependencies
• you want sentence-transformers or langchain integration

When to avoid

• you need state-of-the-art accuracy and can afford transformer inference
• your task depends on contextual token representations
• you need multilingual coverage not offered by available models

Facets

library · maturity active

machine-learning nlp rag search-engine machine-learning python cross-platform static-embeddings sentence-transformers embeddings distillation cpu-inference text-classification retrieval natural-language-processing retrieval-augmented-generation search cpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="MinishLab/model2vec")