- Box unveiled Box Content Retrieval, a new feature built on Nvidia’s Nemotron embedding models, at its developer conference on Nov. 5, 2026.
- Nvidia VP Kari Briski said the system should answer queries four times faster, with 20% fewer errors and 90% fewer tokens used.
- The feature rolls out to Box’s Enterprise Advance customers in early 2027 and relies on Box’s existing metadata and permission controls.
- CEO Aaron Levie said agents can now retrieve relevant document chunks directly instead of reading entire files to answer a query.
What Happened
Box unveiled Box Content Retrieval, a feature that builds Nvidia’s Nemotron embedding system into its core product, at its developer conference on Nov. 5, 2026, TechCrunch (Nov. 5, 2026) reported. Nvidia’s VP of generative AI software, Kari Briski, told TechCrunch she expects queries answered “four times faster, with 20% fewer errors.”
Why It Matters
The feature is an implementation of Nvidia’s Nemotron Embed project, part of the same open-weight model line covered in Nvidia’s Nemotron 3.5 Bolt Packs a Local Voice Agent Into 4GB of VRAM. Nvidia’s open embedding models have topped the RTEB retrieval benchmark as recently as July 2026, giving enterprise vendors like Box a ready-made retrieval layer instead of building embeddings in-house.
Technical Details
Box says the new system cuts token usage by 90% while reducing errors by 20% and speeding query responses roughly fourfold, by sending agents the most relevant document chunks instead of full files. Nemotron’s embeddings map underlying concepts rather than matching keywords, letting the system surface related material even when wording differs, which Box says will help use cases like legal teams searching for related clauses and precedents.
Who’s Affected
Box’s Enterprise Advance customers gain the new retrieval feature when it rolls out in early 2027. The system depends on Box’s existing metadata and permission infrastructure, which the company says makes it difficult for competitors to replicate the feature from scratch using Nemotron alone.
What’s Next
Box Content Retrieval is scheduled to roll out to Enterprise Advance customers starting early 2027, and Nvidia continues positioning its Nemotron Embed models as an enterprise retrieval layer for other software vendors.
