Box announces support for NVIDIA Llama Nemotron Reasoning Models in Box AI Studio

Box is accelerating support for advanced reasoning in Box AI Studio by integrating the new NVIDIA Llama Nemotron reasoning models.

Businesses rely heavily on a wide range of unstructured data – sales contracts, product launch videos, financial statements, marketing images, audio files, and more. Even small mistakes in interpreting this data can lead to significant costs and critical issues. Now more than ever, the need for accurate answers is paramount. Traditional AI models struggle in these scenarios, generating inconsistent, error-prone answers, or miss insights entirely from rich media files.

To tackle these complex challenges, AI agents must be able to accurately process and understand this diverse information, ensuring better decisions and unlocking greater value from their content. We are unlocking even more advanced reasoning capabilities in Box AI. Coming soon, Box AI Studio will support NVIDIA Llama Nemotron reasoning models so that enterprise customers can process and reason over their diverse content to drive powerful agentic workflows across the enterprise.

The next step in advanced reasoning for Box AI Studio

Recent advances in agentic AI are helping organizations outsource simple tasks like automating data entry or text extraction, but imagine if you had the ability to process and reason over virtually all of your organization’s content.

To further help realize that potential, we’re working with NVIDIA Llama Nemotron reasoning models to enhance how Box AI analyzes, connects, and reasons across content, ensuring that even the most challenging queries about an enterprise’s most important information are met with intelligent, context-aware responses.

Box AI’s advanced reasoning with new reasoning models will enable organizations to:

• Bolster legal analysis: Legal firms or teams could use Box AI to evaluate the strength, logical consistency, and completeness of a legal brief after connecting content from multiple contracts, deposition recordings, and video evidence. This can help quickly identify potential weaknesses, gaps, and conflicting info that could expose an organization to risk.
• Streamline research: Life science companies could use reasoning within Box AI to assess insights across clinical trials, scientific publications, experimental video logs, and data visualizations to accelerate the development and rollouts of new drugs or devices.
• Enhance risk management: Banks and wealth management firms could employ Box AI to read market data, client profiles, and regulatory information, helping support robust risk profiles and flag potential vulnerabilities before they escalate. 
• Optimize marketing campaigns: Marketing departments could mine insights from social media data, campaign videos, promo images, and customer feedback reports. By connecting diverse content types, advanced reasoning within Box AI can help enterprises compare emerging trends and potential strategies, enabling more targeted campaigns.
• Accelerate claims processing: Insurance companies could use advanced reasoning AI to evaluate incident reports, customer communications, and historical claims data. This can help validate claims faster, detect fraud early, and reduce settlement times.

 This collaboration with NVIDIA to incorporate reasoning capabilities into Box AI is driven by our belief that advanced logic within agents is essential for tackling the complex, multi-step challenges our enterprise users handle every day. Our approach to AI agents is that they must enable the following:

• Expertise: To handle complex business needs, agents need models that are backed by reinforcement learning and optimization techniques, which help elevate agentic reasoning and planning capabilities used by our enterprise customers.
• Flexibility: Agents must excel at a variety of tasks and make informed decisions to support the diverse needs across any organization in any industry.
• Deployment readiness: Microservices with integration via industry-standard APIs are a must for agents to be deployed within large, global organizations.
• Contextual awareness: By understanding user intent and offering the ability to seamlessly complement other platform capabilities, agents need to deliver precise, actionable insights.
• Trust: For agents to be trusted, they need to be experts in reasoning, secure in how they handle data and safely operate within an enterprise’s existing governance framework, while also being performant in their tasks.

Looking Ahead

While Box AI and Box AI Studio have made remarkable progress, we’re just getting started. We’re excited to be exploring Box AI support for Q&A across multiple file modalities – text, video, images, and audio, at the same time – and NVIDIA NeMo™ Retriever is a key part of this effort. Using Box AI across much more of an organization’s content can widen its aperture, bringing a fuller view of unstructured information. Such a dramatic expansion can enhance various business processes, such as easier discovery to locate a critical segment in a video, fuller classification of images or recordings for compliance, or richer analysis and reasoning across multiple content types for even more powerful results.

 We’re excited about the transformative impact these innovations can have on our customers who rely on Box not just for securely managing their content, but also for running their enterprise. We look forward to deepening our collaboration with NVIDIA and delivering these powerful capabilities so organizations can unleash the true potential of their content. Stay tuned for more updates as we continue to lead the way in intelligent content management, and in the meantime, visit box.com/ai to learn more.

The Box product roadmap and features are subject to change. The development, release and timing of any Box products or features described in this blog are (a) at the sole discretion of Box and (b) subject to certain assumptions regarding available resources, product plans and other information available to Box as of the date of this post. The information in this post is not a commitment, promise or legal obligation to deliver any specific features, code or functionality. Customers should make any purchasing decisions based on Box product features and functionality that are currently available.