On June 24, 2026, Google rolled out a major update to Gemini 3.5 Flash, adding a capability called Computer Use. It lets the model operate a computer's graphical interface directly — moving the cursor, clicking buttons, typing into fields and navigating between applications — so it can carry out multi-step tasks as an autonomous agent rather than only returning text.
What Computer Use does
With Computer Use, Gemini 3.5 Flash can take a goal described in plain language and complete it inside real software: filling out forms, pulling data across web pages, or running a sequence of actions in a browser or desktop environment. It is the same class of agentic interface control already offered by some competitors, now built into Google's fastest stable model rather than a separate research preview.
Why it matters
Flash is Google's speed-and-cost tier, so adding Computer Use here makes agent-style automation cheaper to run at scale than it would be on a top-end reasoning model. For developers building assistants that act on a user's behalf, it lowers the barrier to shipping tools that do work, not just describe it.
What about Gemini 3.5 Pro?
The larger Gemini 3.5 Pro — with the Deep Think reasoning mode and a 2-million-token context window — was announced at Google I/O and originally expected in June, but its wide release has slipped to July 2026 while Google refines it on closed-test feedback. Reports about a "Gemini 2.5 Pro" launching on June 22 are a version-number mix-up: the 2.5 line dates to mid-2025 and is being phased out.