Ettin Reranker Family: 6 CrossEncoder Models 17M–1B, SOTA at Every Size

Hugging Face has released the Ettin Reranker family: six CrossEncoder models ranging from 17M to 1B parameters, achieving state-of-the-art performance at every size class. Built on Ettin ModernBERT encoders, trained on approximately 143 million triples, with the full training recipe released publicly.

1 min read|agenticonsult Intelligence

Ettin Reranker Family: 6 CrossEncoder Models 17M–1B, SOTA at Every Size

Hugging Face has released the Ettin Reranker family, six CrossEncoder reranking models spanning 17M to 1B parameters, each achieving state-of-the-art performance in its size class. Built on Ettin ModernBERT encoder backbones and trained on approximately 143 million triples, the family covers a wide range of inference budget constraints. The full training recipe is published alongside the model weights.

Why It Matters

A SOTA reranker at 17M parameters is deployable in latency-constrained or memory-limited environments where larger rerankers are impractical. The complete training recipe release also gives teams an open path to domain-adapted fine-tuning — unusual for a released SOTA model family.

This breaking-news item was assembled from the cited primary source with AI assistance. It is intended for rapid situational awareness — refer to the original publication for the definitive statement.

Ettin Reranker Family: 6 CrossEncoder Models 17M–1B, SOTA at Every Size

Ettin Reranker Family: 6 CrossEncoder Models 17M–1B, SOTA at Every Size

Why It Matters

Live Intel Feed