With Gemini 3.5 Flash, Google bets its next AI wave on agents, not chatbots

On Tuesday, Google unveiled its latest AI model, Gemini 3.5 Flash, which the company touts as the most advanced to date, particularly in the realms of coding and autonomous AI agents. This announcement was made during the annual I/O developer conference, marking a significant shift in Google's approach to AI from mere conversational tools to powerful agentic systems capable of executing complex tasks with minimal human oversight. The Gemini 3.5 Flash model is designed to autonomously handle coding pipelines, manage research endeavors, and even create an operating system from the ground up, based on internal tests. Koray Kavukcuoglu, DeepMind's chief technologist, emphasized the model's impressive combination of speed and quality, stating that it outperforms its predecessor, the 3.1 Pro model, across various benchmarks, including coding and multimodal reasoning. In fact, 3.5 Flash is reported to be four times faster than other leading models, with an optimized version achieving speeds that are twelve times quicker while maintaining high-quality output. During the I/O event, Google engineer Varun Mohan showcased how multiple agents could collaborate on different components of a project, ultimately coming together to construct a full operating system using Antigravity, Google's dedicated platform for agentic development. This highlights Flash's capability to execute long-running tasks efficiently by allowing several AI agents to operate concurrently. The impact of 3.5 Flash is already evident among Google's partners, including banks and fintech companies, where it is streamlining complex workflows. While the model can operate autonomously for extended periods, Google has designed it to request human input when it encounters situations requiring judgment or decision-making. Additionally, Google plans to release the 3.5 Pro model, which will work in conjunction with Flash. Doshi remarked that 3.5 Pro will serve as a planner and orchestrator, utilizing Flash as various sub-agents for specific tasks. This collaborative dynamic between the two models aims to enhance reasoning power and task efficiency. Currently, 3.5 Flash serves as the default model within the Gemini app and Google Search’s AI mode. The company has also announced new agentic features for Search, enabling users to create and manage AI agents directly on the platform. Furthermore, Gemini Spark, a new personal AI agent designed to assist users in managing their digital lives, will be powered by this advanced model. As Google rolls out these capabilities to a broader audience, it faces increased scrutiny. The company is currently under legal examination following a tragic incident involving a user who experienced severe distress after interacting with Gemini. In response, Google has implemented stronger safeguards within the model, especially concerning sensitive topics, ensuring that Gemini 3.5 Flash is better equipped to handle such inquiries. As of now, Gemini 3.5 Flash is available through Antigravity, the Gemini API, Gemini Enterprise, and within the Gemini app and AI mode in Search, signaling a new era for AI-powered tools that blend autonomy with user engagement.