Box Integrates Nvidia Nemotron for AI Content Retrieval
Box has announced a new content retrieval system powered by Nvidia's Nemotron embedding models, promising faster and cheaper agent queries.
Cloud content management platform Box announced a new feature at its developer conference integrating Nvidia’s Nemotron embedding system directly into its product architecture. Named Box Content Retrieval, the capability is designed to help AI agents process queries more efficiently by utilizing specialized metadata generated by Nemotron.
According to Nvidia Vice President of Generative AI Software Kari Briski, the integration aims to deliver substantial performance gains. Speaking to reporters, Briski noted expectations for queries to resolve four times faster while cutting error rates by 20% and reducing token usage by 90%.
Aaron Levie, CEO of Box, explained that the technology significantly improves search relevance for automated agents. Rather than forcing an agent to parse entire documents sequentially, the system extracts and transmits only the specific text chunks most relevant to a user's prompt or the task at hand.
The Technology Behind the Integration
The feature utilizes Nvidia's Nemotron Embed project, a collection of open-weight embedding models built to enhance automated retrieval systems. While major artificial intelligence labs have often prioritized other areas, Nvidia's open-source model releases have gained attention, with a recent July release achieving top scores on the RTEB benchmark for retrieval quality.
Box highlights the potential utility for professionals managing dense documentation, such as lawyers searching for specific clauses and legal precedents. Because Nemotron embeddings map underlying concepts rather than relying strictly on exact keywords, the tool can surface pertinent legal themes even when different terminology is used in the query.
While the underlying Nemotron models are freely available, Box integrates them with its existing metadata framework and enterprise permission controls. Levie described the combination as packing approximately five distinct technologies into a single API structure.
Rollout and Availability
The new capability is scheduled to roll out early next year and will be available exclusively to Box's Enterprise Advance customers.
What it means for developers
For developers building autonomous agents and retrieval-augmented generation pipelines, the shift toward concept-based embedding models like Nemotron highlights the growing importance of efficient data pre-processing. Handling large volumes of enterprise text requires sophisticated embedding strategies to keep latency and token costs manageable. Developers looking to experiment with various foundational technologies can try top AI models cheaply through one API at https://apixoai.online, allowing them to test and integrate multiple leading architectures without managing separate provider keys.
Source: Box taps Nvidia’s Nemotron for new agent-first content retrieval system — TechCrunch AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
