# We built a new way to train contextual embedding models, which encode each chunk of a document with the whole document in view

**URL:** <https://community.perplexity.ai/t/we-built-a-new-way-to-train-contextual-embedding-models-which-encode-each-chunk-of-a-document-with-the-whole-document-in-view/6248>\
**Category:** Announcements\
**Tags:** api\
**Created:** [September 30, 2026, 8:08pm UTC](https://community.perplexity.ai/t/we-built-a-new-way-to-train-contextual-embedding-models-which-encode-each-chunk-of-a-document-with-the-whole-document-in-view/6248 "2026-09-30T20:08:09Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![pplx-announcements](https://avatars.discourse-cdn.com/v4/letter/p/c37758/32.png) [@pplx-announcements](https://community.perplexity.ai/u/pplx-announcements)\
**Post date:** [September 30, 2026, 8:08pm UTC](https://community.perplexity.ai/t/we-built-a-new-way-to-train-contextual-embedding-models-which-encode-each-chunk-of-a-document-with-the-whole-document-in-view/6248/1 "2026-09-30T20:08:09Z")

</div>

pplx-embed-v2-context-9b-preview sets a new state of the art on ConTEB and turbopuffer’s new, privately held context-bench.

> **[Contextual embedding beyond the gold passage](https://www.perplexity.ai/hub/blog/contextual-embedding-beyond-the-gold-passage?utm_source=forum&utm_medium=referral&utm_campaign=pplx-embed-v2)**
>
> A new training method, model, and benchmark for retrieving answers and their supporting context.

 ![pplx-turbopuffer.jpeg](https://us1.discourse-cdn.com/flex001/uploads/sonar1/original/2X/e/e9752e5ad7b468365ca2b119278dfc5fe98f8c6f.jpeg)