ModernBERT

Evidence (4)

• 7Hugging Face Blog·7/28/2026model_release
  LiquidAI releases LFM2.5 encoders for fast long-context CPU inference

  Hugging Face announced two new encoder models, LFM2.5-Encoder-230M and LFM2.5-Encoder-350M, designed for long-context inference with strong accuracy and low latency on CPU. The models support up to 8,192 tokens, are trained with masked language modeling, and are positioned for use cases such as classification, intent routing, safety filtering, and PII detection.

• 7Hugging Face Blog·5/19/2026model_release
  Hugging Face releases Ettin reranker family

  Hugging Face announced six new Sentence Transformers CrossEncoder rerankers built on Ettin ModernBERT encoders, ranging from 17M to 1B parameters. The release includes the models, training data, and full training recipe, and positions them for retrieve-then-rerank pipelines alongside embedding models.

• 8Hugging Face Blog·5/14/2026model_release
  IBM Granite Releases Multilingual Embedding R2 Models

  IBM has released two Apache 2.0 multilingual embedding models based on ModernBERT: a 97M-parameter compact model and a 311M full-size model. The models support 200+ languages, 32K-token context, and code retrieval across nine programming languages, with strong retrieval benchmark results and Matryoshka support for the larger model.

• 7Hugging Face Blog·3/31/2026research
  Open-source mRNA language models trained across 25 species for $165

  OpenMed described an end-to-end protein engineering pipeline covering structure prediction, sequence design, and codon optimization. Its codon-optimization work found CodonRoBERTa-large-v2 to be the best performer, then scaled the approach to 25 species with four production models trained in 55 GPU-hours.