NOW LET US – AI RAG SaaS Studio TP.HCM
NOW LET US
Digital Product Studio
Back to news
AGENTIC-SYSTEMS...1 min read

Introducing computer use in Gemini 3.5 Flash

Share
NOW LET US Article – Introducing computer use in Gemini 3.5 Flash

Google has natively integrated the 'computer use' capability into Gemini 3.5 Flash, enabling developers to build custom agents that can see, reason, and take actions across browser, mobile, and desktop environments.

Introducing computer use in Gemini 3.5 Flash

Computer use is now a built-in tool supported in Gemini 3.5 Flash, delivering our best performance yet for agentic computer use tasks. Previously only available as a standalone Gemini 2.5 computer use model, computer use is now integrated natively in the main Gemini Flash model. Gemini already excels at function calling and using built-in tools like Search and Maps grounding. With built-in computer use capability, developers can now use 3.5 Flash to reliably build custom agents that can see, reason and take action across browser, mobile and desktop environments. This unlocks improved performance for long-horizon and enterprise automation tasks like continuous software testing and knowledge work across professional applications.

3.5 Flash uses computer use to analyse the Gemini app and return a categorized list of features.

3.5 Flash with computer use audits its own documentation for accessibility issues.

Making computer use safe in 3.5 Flash

To mitigate some of the prompt injection risks for agents operating in live environments, we use targeted adversarial training for computer use in Gemini 3.5 Flash. We’re also releasing two optional enterprise safeguard systems that enable enterprises to:

  • Require explicit user confirmation for sensitive or irreversible actions.
  • Automatically stop tasks if an indirect prompt injection is identified.

Taking a “defense-in-depth” approach, we encourage developers to combine these features with secure sandboxing, human-in-the-loop verification and strict access controls. Additional information on safety measures can be found in our best practices documentation.

We are already seeing customers drive value with computer use. Here’s what some of them have to say:

To start building with computer use today:

Try it now: Test the capabilities in a demo environment hosted by Browserbase.Start building: Dive into our reference implementation and documentation via Gemini API and Gemini Enterprise Agent Platform.

© 2026 Now Let Us. All rights reserved.

Source: Google DeepMind Blog

Advertisement
Ad slot ready: 5887729102

More in this category

NOW LET US Related – Accelerating Skill Assessment in Chess: A Drift-Diffusion-Enhanced Elo Rating System

agentic-systems

Accelerating Skill Assessment in Chess: A Drift-Diffusion-Enhanced Elo Rating System

Researchers have developed DD-Elo, a new chess rating system based on the drift-diffusion model from cognitive neuroscience. By analyzing move-by-move data rather than just match outcomes, DD-Elo updates player ratings much faster and more accurately than the traditional Elo system.

NOW LET US Related – Knowledge-augmented Agentic AI for Mental Health Medication Information Seeking

agentic-systems

Knowledge-augmented Agentic AI for Mental Health Medication Information Seeking

Researchers have developed a knowledge-augmented multi-agent AI framework that integrates regulatory FDA records with patient narratives from Reddit and WebMD, offering a safer and more traceable way to seek mental health medication information.

NOW LET US Related – Agentic Analysis for Agentic Infrastructure: An LLM-Powered Pipeline for Comparative Governance of DAO and Corporate AI Protocols

agentic-systems

Agentic Analysis for Agentic Infrastructure: An LLM-Powered Pipeline for Comparative Governance of DAO and Corporate AI Protocols

Researchers have introduced an LLM-powered comparative pipeline to analyze the governance structures of AI agent protocols, comparing decentralized (DAO) and corporate-led standards.

NOW LET US Related – Geometry-Aware MCTS for Extremal Problems in Combinatorial Geometry

agentic-systems

Geometry-Aware MCTS for Extremal Problems in Combinatorial Geometry

Researchers have proposed a Geometry-Aware MCTS framework to solve complex extremal problems in combinatorial geometry. This new approach overcomes the limitations of traditional RL and Transformer models, establishing new best-known computational results on five out of six tested problems.

NOW LET US Related – Narration-of-Thought: Inference-Time Scaffolding for Defeasible Ethical Reasoning in Large Language Models

agentic-systems

Narration-of-Thought: Inference-Time Scaffolding for Defeasible Ethical Reasoning in Large Language Models

Researchers introduce Narration-of-Thought (NoT), a zero-training inference-time scaffolding that structures LLM reasoning to dramatically improve ethical decision-making and reduce cognitive biases like stakeholder collapse and uncertainty suppression.

NOW LET US Related – Accelerating Returns and the Qualitative Engine for Science

agentic-systems

Accelerating Returns and the Qualitative Engine for Science

While technological progress accelerates exponentially, genuine scientific discovery requires qualitative reasoning that current AI lacks. This paper introduces the Qualitative Engine for Science (QES) to bridge the gap between quantitative execution and human-like conceptual breakthroughs.

EXPLORE TOPICS

Discover All Categories

Deep dive into the specific technology sectors that matter most to you.