We're releasing pplx-embed-v2-late, two late-interaction embedding models that retrieve text, images, and pages with a shared embedding space for cross-model querying.

Both models achieve frontier performance and are publicly available on Hugging Face.

perplexity.ai/hub/blog/multi…

Dense embedding models compress each document into one vector, which loses detail as pages get longer or more visual.

pplx-embed-v2-late keeps a 128-dimensional vector per token and scores with MaxSim, so each query token is matched to its closest token in the document.

The models also embed images and rendered pages directly, so PDFs, slides, and scans are searched without OCR, keeping tables, figures, and layout that text extraction drops.

Both sizes are distilled token by token from one 18B teacher, so they share one embedding space.

A corpus indexed with the 9B model can be searched with 0.6B queries. That lifts ViDoRe v3 from 62.3% to 63.5% with no added query cost.

On Q2D-Web, about 70,000 production-derived search queries over 190M web documents, the 9B and 0.6B models reach 74.8% and 73.6% Recall@1000, against 69.3% for nemotron-embed-8b.