Google Unveils Gemini 3.5: A New Generation of AI Capable of Autonomous, Human-Like Workflow Execution

While previous iterations of AI have largely functioned as conversational assistants, Google has engineered Gemini 3.5 to serve as a high-functioning collaborator. The most significant advancement in this version is its "agentic" nature, which allows the artificial intelligence to independently perform actions to solve real-world problems. Google has inaugurated this series with the 3.5 Flash model, precision-tuned to achieve a synergy of superior intelligence and rapid processing speeds.

Google's Gemini 3.5 version

According to internal testing by Google DeepMind, this new version achieves processing speeds four times faster than current industry-leading models. Empowered by this performance, the AI can autonomously undertake complex, long-horizon tasks—such as financial auditing or managing massive codebases—that previously consumed weeks of human labor. Notably, when paired with the Antigravity framework, Gemini 3.5 Flash can autonomously plan, delegate tasks to sub-agents, and execute complex, multi-step workflows with precision.

One of the most practical applications introduced at the event is Gemini Spark, a personal AI agent powered by the 3.5 Flash architecture. Rather than requiring users to manually manage schedules or scour through disparate data sources, Gemini Spark operates 24/7, processing digital tasks according to specific instructions. This intelligent tool is currently undergoing trials and is slated for a Beta release for Google AI Ultra subscribers in the United States beginning next week.

Beyond its utility for individual users, Google's new model demonstrates superior performance on professional coding benchmarks, such as Terminal-Bench and GDPval-AA, significantly outperforming the preceding 3.1 Pro version. Google’s financial partners have already successfully leveraged this model to automate processes that previously spanned weeks, compressing them into short timeframes while slashing operating costs by over 50%.

Despite its powerful processing capabilities, Google has prioritized safety in the development of Gemini 3.5, strictly adhering to the standards of the Frontier Safety Framework. The company has proactively implemented a series of protective layers to mitigate cybersecurity threats and minimize the generation of harmful content. A notable core technological development is Google's integration of interpretability analysis tools, which allow for rigorous vetting of the AI's internal reasoning processes before any responses are provided to the user.

Starting now, users worldwide can experience the versatility of Gemini 3.5 Flash directly through the Gemini app or via the integrated AI mode on the Google Search platform. Meanwhile, developers have been granted access to the model through the Gemini API on Google AI Studio and Android Studio. A Google representative further disclosed that the premium version, Gemini 3.5 Pro, is currently undergoing internal refinement and is scheduled for a public release next month.