GLiNER

GLiNER is a framework for training and deploying Named Entity Recognition models that supports zero-shot entity extraction on consumer hardware and CPUs.

GLiNER is a framework for training and deploying Named Entity Recognition (NER) models with zero-shot capabilities. It allows users to extract entity types without requiring labeled data or task-specific training. The system supports various architectures including uni-encoder, bi-encoder, and RelEx for joint entity and relation extraction. Models run on standard CPUs and consumer hardware, utilizing optimizations like quantization and torch.compile for efficiency. The framework includes a Ray Serve-based layer for production pipelines requiring high throughput and horizontal scaling. It also offers a growing ecosystem of community projects for C++, Rust, and spaCy integration.

GLiNER at a glance

Best for: Software engineer, Researcher
Runs on: API, self-hosted, open source
Pricing: Free
Released: November 2023
Category: AI coding tool
Website: github.com/urchade/GLiNER

Key features

Zero-shot named entity recognition without training data
Joint extraction of entities and relations
Support for PII detection in over 100 languages
Quantization and compilation for efficient hardware execution
Ray Serve-based serving layer for high-throughput production
Fine-tuning capabilities for domain-specific corpora
Streaming NER and rolling prediction updates
Integration with spaCy, vLLM, C++, and Rust

Use it for

Compliance and PII redaction across documents and data pipelines
Knowledge graph construction via joint entity and relation extraction
Large-scale entity extraction for production document tagging
Domain-specific NER for financial, legal, or biomedical texts
Multi-lingual information extraction

Alternatives to GLiNER

DeepKE — It supports multi-modal, few-shot, and document-level extraction scenarios.
jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
nerfguard-extract — Delivers frontier quality extraction for lower cost and faster speed.
Hugging Face Transformers — It provides a unified API for over 1M+ pretrained model checkpoints on the Hub.
Pagesift — Extracts web data with high memory efficiency and speed for AI agents.
CTranslate2 — Provides fast and efficient inference execution for Transformer models.
LangExtract — It maps every extraction to its exact location in the source text.
Infinity — Achieves 0.1 milliseconds query latency on million-scale vector datasets.
GraphRAG — Enhances LLM reasoning over private data using knowledge graph structures.
Hyper-Extract — Extracts complex knowledge graphs and hypergraphs from documents with one command.
Scikit-LLM — LLM text classifiers with the familiar scikit-learn fit and predict API
Ludwig — Train and fine-tune complex AI models using a single YAML configuration file.

All GLiNER alternatives