pplx-embed-v2-late-9b
Perplexity’s large late-interaction embedder for text and visual document retrieval.
pplx-embed-v2-late-9b is the large late-interaction embedding model from Perplexity, producing 128-dimension per-token vectors for text and images. Perplexity reports the highest average on its domain-specific text retrieval suite (81.3% nDCG@10) and 64.0% on BrowseComp+ (vendor-reported).
- Provider
- Perplexity
- Type
- Embedding model
- Released
- Oct 7, 2026
- Input
- text, image
- Output
- embedding
- Open weights
- Yes
- License
- MIT
Best for
- Research
- Vision
- Enterprise
Strengths
- Searches PDF pages without OCR
- Multi-vector (ColBERT-style) scoring
More from Perplexity
- Decider v1.1 27B — Perplexity. Perplexity’s open-weight multimodal decision model.
- pplx-embed-v1 4B — Perplexity. Perplexity’s web-scale retrieval embedding.
- pplx-embed-v2-late-0.6b — Perplexity. Small late-interaction (ColBERT-style) embedder that searches text and PDF pages.
Tools that use it
- Vespa — Vespa.ai. Open-source engine for search, recommendation and RAG at scale.