What are the key points?

Gemini in Google Sheets achieves 70.48% success rate on the SpreadsheetBench dataset. New beta features allow users to create and edit complex spreadsheets using natural language. Gemini's spreadsheet manipulation performance now rivals human expert ability in real-world scenarios.

Gemini Hits State-of-the-Art Spreadsheet Performance

•Gemini in Google Sheets achieves 70.48% success rate on the SpreadsheetBench dataset.
•New beta features allow users to create and edit complex spreadsheets using natural language.
•Gemini's spreadsheet manipulation performance now rivals human expert ability in real-world scenarios.

Google has announced a significant milestone for Gemini in Google Sheets, claiming state-of-the-art performance that rivals human experts. By integrating advanced generative capabilities directly into the spreadsheet interface, users can now describe complex data tasks in plain language. The system then autonomously executes these requests, ranging from basic formatting to intricate data analysis.

Central to this update is Gemini’s performance on SpreadsheetBench, a public benchmark—essentially a standardized test—designed to evaluate how well AI models can navigate and modify spreadsheets in realistic scenarios. Gemini achieved a 70.48% success rate, a score that surpasses current competitors and moves the needle closer to the proficiency of professional human analysts. This capability allows the AI to understand the logic of a sheet without manual intervention.

These new beta features are part of a broader push to weave AI into the Google Workspace ecosystem. Beyond simple automation, the goal is to provide a collaborative partner that handles the heavy lifting of data management. For students and professionals alike, this means less time spent on manual cell editing and more time focused on interpreting the insights generated from their data.

Google announced that its AI, Gemini, has become a superstar at using Google Sheets. This means instead of hunting for buttons, you can just tell the AI what to do in plain English. The AI then finishes the task on its own, from changing how things look (formatting) to solving tricky math puzzles (data analysis).

To see how good it really is, Gemini took a special standardized test (SpreadsheetBench) used to grade how well AI works with spreadsheets in real life. Gemini got a score of 70.48%, which is currently the best in the world (state-of-the-art). This score is very close to what professional human workers can do. It can understand the logic behind all the numbers and boxes without a human explaining it first.

These new early-version features (beta features) are being added to Google’s apps to help everyone. The goal is to have the AI act like a smart partner that does the boring, hard work (heavy lifting) of managing data. This way, students and adults don't have to spend forever typing into tiny boxes (cells). Instead, they can spend more time learning what the information is actually telling them.