Google has integrated computer use as a native capability in Gemini 3.5 Flash, allowing the model to see, reason, and take action across browser, mobile, and desktop environments natively without separate API calls. Computer use was previously only available as a standalone Gemini 2.5 model but is now built into the main Flash tier. Developers can now use 3.5 Flash via the Gemini API and Gemini Enterprise Agent Platform to reliably build custom agents capable of long-horizon and enterprise automation tasks like continuous software testing, knowledge work, and cross-application process automation.
The integration is significant because Flash (1M token context, $1.50/$9 per million input/output tokens) now replaces the more expensive Pro-tier for agent workloads. Gemini 3.5 Flash already beats Gemini 3.1 Pro on agentic benchmarks (MCP Atlas 83.6%, GDPval-AA 1656 Elo) with 4x the output speed. Adding native computer use to Flash means developers can build screen-aware agents at lower cost and higher speed than competitors. Google has applied targeted adversarial training to mitigate prompt injection risks and is releasing optional enterprise safeguard systems: explicit user confirmation for sensitive actions and automatic task-stopping on detected indirect prompt injection.
For teams building agents: This is the moment to re-evaluate your agent stack. Flash now bundles function calling, retrieval, search grounding, and computer use in one model at a price point 40% cheaper on output than before. The native integration also means no separate network hop for screen interaction—lower latency for browser automation, RPA, and testing workflows. Watch for competitors (Claude, OpenAI) to respond on computer use capability and pricing.