GLiNER 2-XL is here

State of the art classification and NER all in one model

First 1000 requests free

600,000 Monthly Downloads

2400 GitHub Stars

90M+ End Users

Efficiency at Scale

Half the cost, full-scale performance.

Run enterprise-grade NER and classification for $0.625 per 1 M tokens.

Blazing-fast inference.

Average latency ≈ 130 ms per request — built for real-time pipelines.

New capabilities with GliNER-2

Named Entity Recognition

Identify people, companies, locations, and concepts in any text — instantly.
GLiNER 2-XL uses zero-shot transfer and contextual embeddings to detect new entities with human-level precision, without retraining.

Text Classification

Understand intent, sentiment, and topic at scale.
Classify messages, documents, or conversations into fine-grained categories that evolve with your users — powered by adaptive, self-learning agents.

Structured Extraction

Turn unstructured text into clean, actionable data.
Automatically extract key fields like amounts, addresses, or events, and map them into your database or workflow — ready for analytics, automation, or compliance.

GliNER-2-XL NER Benchmark Results


Dataset

GPT-5

GLiNER-M

GLiNER-2-XL

Parameters

-

209M

1B

Hosting

-

Self-host

Fastino hosted (A100)

Latency

7000-28000 ms

160 ms

130 ms

CrossNER - AI

0.547

0.518

0.573

CrossNER - Literature

0.561

0.597

0.572

CrossNER - Music

0.736

0.694

0.717

CrossNER - Politics

0.632

0.686

0.675

CrossNER - Science

0.518

0.581

0.623

CrossNER - Average

0.599

0.615

0.636


Community & support

Join the community

Join our active community on Discord.

Need help?

Get in touch with our support team.