Vectorium adds RaBitQ
A post announcing the addition claims up to 30× compression with no training, plus speeds 1.4–2.2× faster than the original implementation.
TLDR
The announcement says RaBitQ is now in Vectorium, supporting 1, 2, 4 or 8 bits per component. It claims up to 30× compression without training and speeds 1.4–2.2× faster than the original implementation. For neural embeddings, it also claims RaBitQ beats PQ “on every axis,” with the gap growing as dimensionality increases.
Combined views
38
1 Source, first seen 16d ago