Box eval: Gemini 4 Argon gains 14 points on enterprise agent tasks
- Source
- x.com
- Date
At Box, we've been testing Gemini 4 Argon on our Box AI Complex Work eval for the past week. It's a huge upgrade over Gemini 3.1 Pro, jumping by 14 percentage points across all our tests, and firmly at the frontier. In our evaluation, we give the Box Agent (using Gemini 4 Argon) a variety of complex enterprise tasks involving enterprise documents in a range of industries. These tasks approximate the kind of hard, time-consuming work that knowledge workers do every day with their enterprise content. Gemini 4 is a huge leap on effectively every dimension of this work. Gemini 4 is far more capable on math, logic, analytics, and deep knowledge of a variety of business domains. It also works far harder to complete tasks fully without stopping short of the desired result. This is incredibly important for complex enterprise workflows that depend on robust task execution. Here are some example wins that we saw to put the improvements in perspective: * Financial services (+18 points): a funding-gap analysis for a family-owned company, where the brief spelled out every step of the projected statements, down to the interest on the bank loan and the overdraft for each year. Gemini 4…
Context
Box's Aaron Levie reports that after a week of testing, Google's Gemini 4 Argon scored 14 percentage points higher than Gemini 3.1 Pro across Box's internal Box AI Complex Work (x.com). The eval has the Box , running on the model under test, work through complex tasks on enterprise documents from several industries, meant to approximate hard, time-consuming knowledge work. The post gives only the size of the improvement in text, not absolute scores, and it is Box's own rather than an independent one.
Levie's per-industry examples illustrate the differences. In a financial-services funding-gap analysis (up 18 points), Argon built the projected statements line by line as the brief specified, while Gemini 3.1 Pro folded the loan interest into a single figure. In a healthcare care plan (up 25 points), both models spotted a dangerous potassium level and a drug interaction, but only Argon removed the offending drugs from its final regimen so the plan matched its own diagnosis. On a technology task (up 37 points), Argon pulled the correct month's figures for two regional markets, while 3.1 Pro often pulled wrong rows, once reporting retention above 100%. Levie says the new model is stronger at math, logic and business-domain knowledge, and works harder to finish tasks instead of stopping short. Box says Argon will soon be available in Box AI Studio for building custom Box Agents.
- eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
Checking sign-in…
Loading comments…