Qdrant Tests Perplexity's Visual Document Embedding Model

Qdrant independently benchmarked Perplexity's new visual document embedding model pplx-embed-v2 across 11,609 pages from 4 ViDoRe v3 datasets. With 1-bit compression, the model shrinks to 1/21 of its size while retaining 96.5% recall.

2026-10-16 ~ 2026-10-17 · 2 related posts

• Qdrant tests Perplexity's visual PDF embeddings: 21x smaller vectors, 99.9% after rescoring — jrxna · 2026-10-16
• Qdrant tests Perplexity's PDF embeddings: 21× smaller vectors, 96.5% recall kept — jrxna · 2026-10-17