urchade/GLiNER

Health v2 · maintenance only

Adoption not part of the score

3997 stars · 317 forks observed · 2026-10-02

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

GLiNER is a Python framework for training and deploying lightweight, zero-shot Named Entity Recognition models that can extract arbitrary entity types from text without labeled data. It also supports streaming NER, joint entity and relation extraction, and multi-task token classification, running efficiently on CPUs and consumer hardware.

Use cases

• extract named entities from text without training data
• detect PII in documents
• perform zero-shot NER with custom entity types
• extract entities and relations jointly
• run NER on CPU without a GPU
• stream entity recognition over incremental text
• fine-tune a small NER model on my own labels

When to choose

• you need flexible entity types without training a task-specific model
• you want LLM-like extraction flexibility at a fraction of the cost and size
• you need to deploy NER on CPU or edge hardware with ONNX/INT8 support

When to avoid

• you need a fixed set of predefined entity types with maximum accuracy from a classic NER model
• you need full generative LLM capabilities beyond entity extraction
• you are not working in the Python/PyTorch ecosystem

Facets

library · maturity active

nlp machine-learning llm-inference machine-learning artificial-intelligence python cross-platform named-entity-recognition zero-shot information-extraction relation-extraction pii-detection transformers onnx token-classification natural-language-processing gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="urchade/GLiNER")