The Gemini 3.5 Flash model now natively supports computer use, a capability that had previously been offered as a standalone Gemini 2.5 computer use model. The integration aims to improve agentic computer tasks, particularly for long‑horizon and enterprise automation scenarios such as continuous software testing and knowledge work across professional applications.
Gemini has already been strong at function calling and using built‑in tools like Search and Maps for grounding. With computer use integrated into the main Flash model, developers can build custom agents on 3.5 Flash that can see, reason, and take actions across browser, mobile, and desktop environments more reliably.
What the built-in computer use can do
- 3.5 Flash can analyze the Gemini app and return a categorized list of features using computer use.
- The model can audit its own documentation for accessibility issues.
These capabilities are intended to help enterprise workflows that require repeatable, long‑running or multi‑application automation.
Access and developer resources
Developers and enterprises can access computer use in 3.5 Flash via the Gemini API and the Gemini Enterprise Agent Platform. A demo environment hosted by Browserbase is available for testing, and a reference implementation and documentation are provided to help teams start building.
Making computer use safe in 3.5 Flash
To reduce prompt injection risks for agents operating in live environments, Gemini 3.5 Flash uses targeted adversarial training for computer use. In addition, two optional enterprise safeguard systems are offered:
- Require explicit user confirmation for sensitive or irreversible actions.
- Automatically stop tasks if an indirect prompt injection is identified.
Google recommends a defense‑in‑depth approach: combine these safeguards with secure sandboxing, human‑in‑the‑loop verification, and strict access controls. More detailed safety guidance is available in the best practices documentation.
Customer adoption and next steps
Google reports that customers are already driving value with computer use. To try the feature now, users can test it in the Browserbase‑hosted demo environment, or begin building using the reference implementation and documentation available through the Gemini API and Gemini Enterprise Agent Platform.
