Box
A U.S.-based cloud content management and file sharing service company that provides enterprises with secure cloud storage, collaboration, and content management solutions.
Core Facts
Timeline (last 90 days)
The knowledge work benchmark was built by the cloud content management platform Box, specifically to evaluate the comprehensive capabilities of AI models in real enterprise scenarios, covering tasks such as data extraction, cross-document association, law, life sciences, and healthcare.
Engineering departments are almost always the first group in a company to embrace AI agents, because engineers' work output is inherently verifiable (whether the code runs, whether tests pass, whether the build succeeds are all immediate feedback).
Aaron Levie is the CEO of Box.
Muse supports work connectors including GitHub, Notion, Granola, and Box.
The Box team applied Opus 5.5 to enterprise knowledge work, achieving 63% token savings, 42% reduction in redundancy, and 30% speed improvement compared to Opus 5, with accuracy gains of 39% on financial services tasks and 65% on cloud cost analysis.
Gemini 4 Argon's maximum output has been increased from 64,000 tokens to 1 million (referring to the output limit, not the context window)
An interviewee from Box said that complex accounting in the media and entertainment sector is one of his favorite use cases, with Astra correctly handling tax incentives and avoiding double-counting of the 20% incentive.
All Facts (7)
Gemini 4 Argon's maximum output has been increased from 64,000 tokens to 1 million (referring to the output limit, not the context window)
80%UnverifiedThe knowledge work benchmark was built by the cloud content management platform Box, specifically to evaluate the comprehensive capabilities of AI models in real enterprise scenarios, covering tasks such as data extraction, cross-document association, law, life sciences, and healthcare.
50%UnverifiedEngineering departments are almost always the first group in a company to embrace AI agents, because engineers' work output is inherently verifiable (whether the code runs, whether tests pass, whether the build succeeds are all immediate feedback).
50%UnverifiedAaron Levie is the CEO of Box.
50%UnverifiedMuse supports work connectors including GitHub, Notion, Granola, and Box.
50%UnverifiedThe Box team applied Opus 5.5 to enterprise knowledge work, achieving 63% token savings, 42% reduction in redundancy, and 30% speed improvement compared to Opus 5, with accuracy gains of 39% on financial services tasks and 65% on cloud cost analysis.
50%UnverifiedAn interviewee from Box said that complex accounting in the media and entertainment sector is one of his favorite use cases, with Astra correctly handling tax incentives and avoiding double-counting of the 20% incentive.
50%