Read full article →

Google bakes computer control directly into Gemini 3.5 Flash, letting the model see and operate your screen

Summary

Gemini 3.5 Flash Gets New Powers: Control Your Computer HOMEPAGE: Google's AI model Gemini 3.5 Flash can now control your computer, see your screen, and even operate mobile devices on its own.
This powerful tool can be used for software testing and office automation.
SUMMARY: Google has upgraded its AI model Gemini 3.5 Flash to include "Computer Use" capabilities, allowing it to operate computers, browsers, and mobile devices independently.
This advancement brings Gemini 3.5 Flash on par with GPT-5.5 in the OSWorld benchmark, scoring 78.4.
Developers can now use the Gemini API to build agents for software testing and office automation.
The update demonstrates the model's growing ability to interact with and control the physical world.
WHY IT MATTERS: As AI continues to advance, we're seeing more models like Gemini 3.5 Flash that can perform complex tasks on their own.
This trend has significant implications for industries like software development, customer service, and office work.
Everyday people will notice this shift in their daily interactions with technology, and it's essential to understand how AI is changing the world.
EXPLANATION: To understand this story, let's break down a few key terms: 1.
API (Application Programming Interface): Think of an API like a set of instructions that allows different software programs to communicate with each other.
In this case, the Gemini API allows developers to build agents that can interact with Gemini 3.5 Flash.
Benchmark: A benchmark is a standard test used to measure the performance of a model or system.
In this story, the OSWorld benchmark is used to compare Gemini 3.5 Flash's capabilities with those of GPT-5.5.
Agent: An agent is a software program that can perform a specific task or set of tasks.
In this case, developers can use the Gemini API to build agents that can control Gemini 3.5 Flash and perform tasks like software testing or office automation.
These terms might seem complex, but they're essential to understanding how AI models like Gemini 3.5 Flash are being developed and used.

Google bakes computer control directly into Gemini 3.5 Flash, letting the model see and operate your screen

Summary

Databricks’ former AI chief thinks he can cut AI’s power bill by 1,000x