GTE Reranker
Alibaba CloudThe reranking half of Alibaba's GTE family: two Apache-2.0 cross-encoders that score query-document pairs for a retrieval pipeline rather than emit vectors. gte-reranker-modernbert-base is a 149M English model on a ModernBERT base reporting BEIR 56.19, LoCo 90.68 and CoIR 79.99 at 8,192 tokens, and gte-multilingual-reranker-base is a 306M encoder-only reranker covering more than 70 languages at the same context, which the mGTE paper positions as roughly ten times faster than the decoder-based alternatives of its generation.
Added 2026-09-11 for the embeddings_retrieval roster, splitting the GTE cross-encoders out from the sibling gte product, which covers the embedders. Scope is the two reranker checkpoints in the Alibaba-NLP account and nothing else; the gte-rerank service on Alibaba Cloud is a separate hosted product the cards note is not identical to the open weights.
Openness
3 high confidence- weights
- open(Apache-2.0 ungated safetensors for both cross-encoders, with ONNX builds beside the ModernBERT one)
- data
- closed(both cards defer the training mixture to the mGTE paper and release no corpus)
- code
- partial(inference usage through transformers, TEI and Infinity ships on the cards
- license
- Apache-2.0(OSI)
The same resolution as the gte embedders, on the same evidence: an OSI license over ungated weights, with the corpus and the training run left in the mGTE paper. No reranker SKU carries a more restrictive license, so nothing pulls the line down.
- https://huggingface.co/Alibaba-NLP/gte-multilingual-reranker-base/raw/main/README.md recorded 2026-09-11
Front matter reads license: apache-2.0. Evaluation is a figure with detail deferred to the mGTE paper (arXiv 2407.19669), no corpus is released, and the usage sections cover transformers, Infinity and TEI inference only.
- https://huggingface.co/Alibaba-NLP/gte-reranker-modernbert-base/raw/main/README.md recorded 2026-09-11
States the gte-modernbert series follows the previous GTE training scheme with ModernBERT as the base and refers readers to the mGTE paper for training details; the usage sections cover transformers, Transformers.js and TEI inference only.
- https://huggingface.co/api/models/Alibaba-NLP/gte-multilingual-reranker-base recorded 2026-09-11
license: apache-2.0; gated: false; private: false; model.safetensors present.
- https://huggingface.co/api/models/Alibaba-NLP/gte-reranker-modernbert-base recorded 2026-09-11
license: apache-2.0; gated: false; private: false; model.safetensors and an onnx/ folder present.
Adoption
4 high confidenceAdoption is measured as Hugging Face downloads across GTE's two reranker checkpoints, most of them for the English-only ModernBERT cross-encoder. The GTE embedding checkpoints are the sibling gte product and are excluded here.
- https://huggingface.co/api/models/Alibaba-NLP/gte-multilingual-reranker-base recorded 2026-09-11
downloads: 372525 for Alibaba-NLP/gte-multilingual-reranker-base.
- https://huggingface.co/api/models/Alibaba-NLP/gte-reranker-modernbert-base recorded 2026-09-11
downloads: 2638520 for Alibaba-NLP/gte-reranker-modernbert-base.
Capability
3 high confidenceGTE Reranker's ModernBERT cross-encoder scores 56.19 BEIR nDCG@10, a step below Jina Reranker's 61.94 on the same benchmark, and it is English-only where the frontier rerankers are multilingual. The multilingual sibling covers 70-plus languages but is an older generation model; the GTE embedding checkpoints are scored separately, and this reading covers only the reranking half of the family.
- https://huggingface.co/Alibaba-NLP/gte-multilingual-reranker-base/raw/main/README.md recorded 2026-09-11
Describes a 306M encoder-only reranker supporting over 70 languages at 8192 tokens, claiming state-of-the-art multilingual retrieval among rerankers of similar size, with results detailed in the mGTE paper.
- https://huggingface.co/Alibaba-NLP/gte-reranker-modernbert-base/raw/main/README.md recorded 2026-09-11
Model table gives gte-reranker-modernbert-base BEIR 56.19, LoCo 90.68 and CoIR 79.99 at 149M parameters, 8192 maximum sequence length and primary language English.
Verified 2026-09-11