gte-reranker-modernbert-base
Alibaba-NLP/gte-reranker-modernbert-base
A popular open reranker model, with 2.7M downloads a month. gigarouter benchmarks and hosts it as an OpenAI-compatible API.
about this model
The gte-reranker-modernbert-base model from Alibaba's Tongyi Lab is a text reranker built on the ModernBERT encoder-only foundation. With 149M parameters and support for up to 8,192 input tokens, it is designed to reorder document passages by relevance to a query, optimizing retrieval pipelines.
Key strengths
- Competitive reranking performance on BEIR (56.19), LoCO (90.68), and COIR (79.99).
- Long-context support (8,192 tokens) enables reranking of lengthy documents and code.
- Efficient size (149M parameters) balances speed and accuracy for production use.
Benchmark highlights
The following table compares the reranker against its embedding counterpart on key benchmarks from the original evaluation:
| Model | Type | BEIR | LoCo | COIR |
|---|---|---|---|---|
| gte-reranker-modernbert-base | reranker | 56.19 | 90.68 | 79.99 |
| gte-modernbert-base | embedding | 55.33 | 87.57 | 79.31 |
On the LoCO long-document retrieval benchmark, the reranker achieves an average score of 90.68, outperforming the base embedding model (87.57) and demonstrating strong capability on tasks like QMSum, SummScreen, and Qasper. Its BEIR score of 56.19 and COIR score of 79.99 reflect robust general and code retrieval reranking.
Gigarouter hosts this model as a managed, OpenAI-compatible API — no infrastructure setup required. Send a query and a list of documents, receive relevance scores for each pair.
We're benchmarking and onboarding gte-reranker-modernbert-base as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.