Perplexitypplx-embed v1
pplx-embed-context-v1-0.6b
This model embeds document chunks with their surrounding context.
What the release provides
The Qwen3-derived model uses diffusion continued pretraining and mean pooling. It outputs 1024-dimensional vectors with a 32K context. INT8 vectors are unnormalized and require cosine similarity. MIT terms apply.pplx-embed-context-v1-0.6b official model card
How to use it well
Practical guidance. Keep text segmentation consistent between indexing and evaluation. Do not compare vectors from unrelated embedding models without an explicit shared-space guarantee.
Hosted access
The embeddings API lists this exact model identifier.Perplexity current embeddings API
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API, Downloadable |
| Model identifiers | perplexity-ai/pplx-embed-context-v1-0.6b |