Bekko Embedding：多语言检索模型能有多小？
原标题：Bekko Embedding: how small can a multilingual retrieval model be?
AI 摘要
Hugging Face 博客介绍了 Bekko Embedding，两个多语言检索模型 bekko-embedding-v1-a8m 和 a25m，分别仅有 7.67M 和 24.93M 活跃参数，在 MMTEB Multilingual v2 的 18 个检索任务上平均分达 56.2 和 57.5，超越多语言 e5 和 BGE-M3。模型通过剪枝 mmBERT-small 并训练 11 亿多语言对得到，支持 100+ 语言、8192 输入、Matryoshka 截断，MIT 许可，可在浏览器中运行。文章强调活跃参数比总参数更能反映推理成本，并指出小模型在本地检索场景（如编码代理）中的价值。
正文节选
What interested me, though, was the opposite of that push for scale and quality — the approach of making the model as small as possible. How far down can you go while keeping quality that is actually usable for real retrieval? Bekko Embedding is where I landed — two models, bekko-embedding-v1-a8m and bekko-embedding-v1-a25m, released under the MIT license and shortened below to bekko-a8m and bekko-a25m. The smaller one has just 7.67M active parameters. Active parameters are the weights used in t