Google Embeddings 2 leads retrieval benchmarks but lags in speed

A new paper benchmarks Google Embeddings 2 (GE2) against several open-source models for multilingual dense retrieval and RAG systems. GE2 outperformed all competitors on accuracy metrics across multiple datasets, including BEIR and an Italian RAG corpus. However, GE2 demonstrated significantly higher latency compared to open-source alternatives, making models like mE5-L more suitable for applications with strict latency requirements. AI

Summary written by gemini-2.5-flash-lite from 1 sources. How we write summaries →

IMPACT GE2 sets a new SOTA for retrieval accuracy but highlights the trade-off between performance and latency, guiding choices for RAG systems.

RANK_REASON Research paper evaluating AI model performance on specific benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• paper
• product