Saved in:
Bibliographic Details
Main Authors: Lansiaux, Edouard, Simonet, Antoine, Wiel, Eric
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.24793
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918379342790656
author Lansiaux, Edouard
Simonet, Antoine
Wiel, Eric
author_facet Lansiaux, Edouard
Simonet, Antoine
Wiel, Eric
contents We present SwiftEmbed, a production-oriented serving system for static token embeddings that achieves 1.12\,ms p50 latency for single-text requests while maintaining a 60.6 MTEB average score across 8 representative tasks. Built around the open-source Potion-base-8M distilled model from MinishLab and implemented in Rust, the system delivers 50,000 requests per second through static embedding lookup, mean pooling, and zero-copy IEEE754 binary serialization. Evaluation demonstrates exceptional duplicate detection performance (90.1% AP) and strong semantic similarity (76.1% Spearman correlation). Performance relative to Sentence-BERT is task-dependent: robust for deduplication and similarity workloads (89--100%), substantially lower for classification and complex retrieval tasks (75%). Domain-specific performance ranges from 75% to 131% of a GloVe-840B baseline. The system targets real-time embedding applications where sub-5\,ms latency is operationally critical and where full transformer inference is not feasible.
format Preprint
id arxiv_https___arxiv_org_abs_2510_24793
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SwiftEmbed: Ultra-Fast Text Embeddings via Static Token Lookup for Real-Time Applications
Lansiaux, Edouard
Simonet, Antoine
Wiel, Eric
Computation and Language
Artificial Intelligence
We present SwiftEmbed, a production-oriented serving system for static token embeddings that achieves 1.12\,ms p50 latency for single-text requests while maintaining a 60.6 MTEB average score across 8 representative tasks. Built around the open-source Potion-base-8M distilled model from MinishLab and implemented in Rust, the system delivers 50,000 requests per second through static embedding lookup, mean pooling, and zero-copy IEEE754 binary serialization. Evaluation demonstrates exceptional duplicate detection performance (90.1% AP) and strong semantic similarity (76.1% Spearman correlation). Performance relative to Sentence-BERT is task-dependent: robust for deduplication and similarity workloads (89--100%), substantially lower for classification and complex retrieval tasks (75%). Domain-specific performance ranges from 75% to 131% of a GloVe-840B baseline. The system targets real-time embedding applications where sub-5\,ms latency is operationally critical and where full transformer inference is not feasible.
title SwiftEmbed: Ultra-Fast Text Embeddings via Static Token Lookup for Real-Time Applications
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2510.24793