Model WarsJune 25, 2026via The Decoder
Google bakes computer control directly into Gemini 3.5 Flash, letting the model see and operate your screen
Why it matters
Google's integration of native computer-use capabilities into Gemini 3.5 Flash represents a significant capability milestone that closes the gap with OpenAI's leading models on agentic reasoning tasks. This democratizes access to autonomous agent building across a broader developer base, shifting the competitive landscape in model capabilities.
Key signals
- Gemini 3.5 Flash now has native Computer Use capability
- Scores 78.4 on OSWorld benchmark
- Performance parity with GPT-5.5 on computer control tasks
- Available via Gemini API for developers
- Use cases: software testing automation, office automation agents
- Models can see and operate screens, browsers, and mobile devices autonomously
The hook
78.4 on OSWorld. Gemini 3.5 Flash just matched GPT-5.5 on computer control—and Google put it in the hands of every developer.
Google has integrated "Computer Use" directly into Gemini 3.5 Flash, letting the model operate computers, browsers, and mobile devices on its own. On the OSWorld benchmark, it scores 78.4, putting it on par with GPT-5.5. Developers can use the Gemini API to build agents for software testing or office automation.