gte-multilingual-reranker-base

The gte-multilingual-reranker-base model is a dedicated reranker from the GTE model family, optimized for multilingual text retrieval. Built on an encoder-only transformer architecture, it delivers inference speeds up to 10x faster than decoder-only rerankers while requiring significantly less hardware. The model handles input lengths up to 8,192 tokens and supports over 70 languages.

Key Features

• High Performance: Achieves state-of-the-art results on multilingual retrieval tasks compared to similar-sized reranker models.
• Efficient Architecture: 306M parameters, designed for low-latency, cost-effective deployment.
• Long-Context Support: Processes documents up to 8,192 tokens.
• Broad Language Coverage: Supports 70+ languages for cross-lingual reranking.

Performance

Evaluated on multiple text retrieval benchmarks, the model demonstrates strong reranking accuracy. The chart below summarizes results across diverse datasets:

Detailed experimental results and comparisons are available in the associated paper.

As a managed API on gigarouter, this reranker can be integrated via a single endpoint, eliminating infrastructure overhead.