Alibaba's Tongyi Lab, DAMO Academy, released the GTE embedding model family.

GTE is a family of general text embedding models, including multilingual and long-context variants, used for semantic search and retrieval.

Hosted on Hugging Face under the Alibaba-NLP and thenlper organizations, with training code on GitHub.

The original GTE models were released in 2023, with the gte-Qwen2 and gte-modernbert variants following through 2024 and 2025.

It provides strong, permissively licensed multilingual embeddings that rank highly on the MTEB leaderboard for retrieval tasks.

Runs through the Hugging Face sentence-transformers or transformers libraries.

Released under the Apache 2.0 license by Alibaba; fully open weights with no gating.