e-means: a faster k-means, now in FAISS 1.16.0
e-means is a published replacement for Lloyd’s k-means. It is in FAISS 1.16.0 and is meant to run on ordinary servers and local hardware.
A public post on 11 November 2026 pointed to e-means, a clustering algorithm now in FAISS v1.16.0. The post notes faster search and AI answers, lower RAM and VRAM use, availability now, and use on standard servers and local hardware without expensive upgrades.
The independent project page and the announcement thread are the source for the figures below. The excerpts are quoted verbatim.
“I’m excited to open-source e-means: a modern, robust k-means algorithm developed at Meta FAIR. A replacement of Lloyd's algorithm, e-means delivers a ~10× speedup in both epochs and wall-clock time while achieving lower loss.”
From the opening of the e-means project page, e-means.github.io.
“Across a large range of scales and datasets, e-means reaches a test loss below Lloyd's at 25 epochs, in 1.2 to 3 epochs. Measured end to end, that is a 10–45× wall-clock reduction on GPU against the state-of-the-art Lloyd implementation (Flash-KMeans [2]), and 8–15× on CPU against Faiss [1]. Speed and quality usually trade against each other in k-means, while e-means achieves both better speed and loss.”
From the section “An order of magnitude less epochs” on e-means.github.io.
“Introducing e-means, an algorithm beating k-means in 10x fewer epochs. With order of magnitude wall-clock speedups, e-means is released in FAISS and needs no hyperparameter tuning beyond k. It is robust to various scales and datasets.”
From the announcement post by Swann Bessa, 10 November 2026.
Swann Bessa et al., e-means: outperforming k-means (in one-tenth the epochs). Project page: e-means.github.io. Paper: arXiv:2611.09501.
Project page · Paper on arXiv · Code in FAISS · Post on X · @swann_bessa · All news