First look: Claude 4 and Box AI

Last week, Anthropic launched its Claude 4 series, including the Sonnet and Opus models, bringing remarkable advancements in coding and development capabilities to the AI landscape. While maintaining strong performance on enterprise use cases and content tasks, this leap forward in technical proficiency makes Claude 4 a uniquely powerful tool for organizations looking to innovate faster with Box AI.

Our Box AI Enterprise Eval confirms that Claude 4 models remain highly competitive for demanding enterprise Q&A tasks, performing near parity with its predecessor, Claude 3.7 Sonnet. However,its potential for developers and technical workflows represents the most significant evolution.

A new benchmark for coding and development

Claude 4 represents a major step forward for AI-assisted development. These models exhibit significant gains in understanding, generating, and debugging code across various programming languages. This enhanced proficiency means developers can:

• Accelerate code generation: Draft functions, build scripts, and create boilerplate code more quickly and accurately.
• Improve debugging: Identify errors, understand complex code blocks, and receive intelligent suggestions for fixes.
• Enhance technical documentation: Leverage AI to understand or even generate documentation from codebases, bridging gaps in knowledge transfer.
• Build smarter AI agents: Create more sophisticated agents that can interact with technical systems, APIs, and code repositories directly.

Claude 4 in action: Automating financial analysis with code generation

We tested Claude 4 by using the Box AI API to analyze ten complex 10-K financial reports and extract key financial data. We used a code editor to prompt Claude 4 to generate Python code that dynamically fetches all file IDs from a specific Box folder, replacing the manual list. We then integrated this snippet and ran the script using Box AI, powered by Claude 4.

In under two minutes, it processed all ten reports, accurately extracting the requested company names, revenues, metrics, and highlights. This example showcases how Claude 4's coding skills, combined with Box AI's content intelligence, can build powerful, automated workflows, saving valuable time for analysts dealing with complex data.

Understanding enterprise content

Powerful coding needs a strong content foundation, as developers constantly interact with documentation, specifications, and project plans within Box. Our Box AI Enterprise Eval confirms Claude 4's continued strength in Q&A and extracting information from content, ensuring it understands and can synthesize this vital context. We found it is highly adept when asked to pull specific details from single documents, and it maintains competitive, reliable performance when synthesizing information across multiple sources or extracting information from enterprise documents. This ensures you don't sacrifice content intelligence for coding power; with Claude 4 in Box AI, you get both, allowing you to build workflows that seamlessly blend structured and unstructured data with code.

Developer-centric use cases

The combination of advanced coding and robust content skills unlocks powerful new use cases. Here are just a few examples of how organizations can leverage Claude 4's unique capabilities within Box:

• Custom engineering agents: Create sophisticated agents that support your engineering teams by referencing technical documents in Box, pulling real-time data from systems like Jira, and finding solutions on platforms like Stack Overflow.
• Technical support bots: Build a support agent that understands your technical manuals and can help users troubleshoot by analyzing code snippets they provide.
• Automated code review: Implement workflows where Claude 4 analyzes code pushed to a repository (stored in Box) against security policies (also in Box).
• Legacy system migration: Use Claude 4 to help understand old codebases and assist in translating them to modern languages or platforms.

Start building today

Claude 4 Sonnet and Opus offer a powerful new dimension to the Box AI platform, especially for organizations with significant development needs. If your team writes code, builds applications, or manages technical infrastructure, Claude 4 provides an unprecedented opportunity to boost productivity and innovation. Start leveraging these capabilities today: Claude 4 Sonnet and Opus are now available in Box AI Studio and through the Box AI API.