---
id: 20260827-T0-02
title: "Giga-Embeddings：100亿参数MoE文本嵌入模型，吞吐量更高"
title_en: "Giga-Embeddings: 10B MoE Encoders Boost Text Embedding Throughput"
url: https://ai.daily.yangsir.net/daily/20260827-T0-02
issue_date: 2026-08-27
publish_date: 2026-08-26T04:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2608.23806
---

# Giga-Embeddings：100亿参数MoE文本嵌入模型，吞吐量更高

arXiv发布Giga-Embeddings，一种文本嵌入模型系列，将强检索质量与高效服务结合。其最大模型为稀疏100亿参数MoE编码器，每token仅激活约18亿参数，在保证检索精度的同时提升了服务吞吐量。该系列模型适用于大规模文档检索和语义搜索场景。

## English Version

**Giga-Embeddings: 10B MoE Encoders Boost Text Embedding Throughput**

arXiv released Giga-Embeddings, a text embedding family balancing strong retrieval quality with efficient serving. Its largest model is a sparse 10B-parameter MoE encoder with ~1.8B active per token, improving throughput without sacrificing accuracy. Ideal for large-scale document retrieval and semantic search.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2608.23806)

**详情页**：https://ai.daily.yangsir.net/daily/20260827-T0-02

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*