pplx-embed-v2-context-9b-preview

Use it to create context-aware document and query embeddings for retrieval-augmented generation systems.

Availability

Preview release on Hugging Face

Pricing model

Not established

Deployment

Load with Transformers and run on a selected device

Privacy

Not established

Licence

MIT

• The model encodes document chunks together so each chunk's embedding reflects its surrounding context. ^[1]

  A document is passed as a list of chunks; the chunks are encoded together, so each chunk's embedding reflects its surrounding context, and one embedding is returned per chunk.

• It returns one embedding per document chunk. ^[1]

  A document is passed as a list of chunks; the chunks are encoded together, so each chunk's embedding reflects its surrounding context, and one embedding is returned per chunk.

• The model has 2048 dimensions and supports Matryoshka dimensions of 1024 and 2048. ^[1]

  The model was trained with Matryoshka losses at **1024** and **2048** dimensions.

• This is a preview release, and weights, embeddings, and the interface may change in later versions without backward compatibility. ^[1]

  Weights, embeddings, and the interface may change in later versions without backward compatibility, so embeddings produced with this preview should not be mixed with embeddings from a future release.

• Queries should be encoded with encode_queries and document chunks with encode. ^[1]

  > Queries and documents are encoded with **different methods**: use `encode_queries` for queries and `encode` for document chunks.

• Backburner leads 25 AI developments in week 40, 2026 Weekly brief 2026-W40
• Perplexity introduces a decision model fine-tuned from Qwen3.8-27B Daily brief 2026-10-01