Google is taking its AI platform to the next level alongside Gemini Omni with Gemini 3.5 and is increasingly focusing on so-called agentic capabilities. The new model family combines high computing power with the ability to independently plan and execute complex tasks over longer periods of time. It starts with Gemini 3.5 Flash, which, according to Google, has been specially optimized for coding, automation and intelligent AI agents. In addition to significantly higher speed, the company promises better multimodal capabilities and practical applications for developers, companies and end users.
- Gemini 3.5 Flash combines high AI performance with autonomous agent functions
- Focus on coding, multi-step workflows and complex long-term tasks
- Four times faster output speed than other Frontier models
- Integration with Gemini app, Google Search, AI Studio and enterprise platforms
Gemini 3.5 Flash is designed to make AI agents significantly more powerful
With Gemini 3.5 Flash, Google is pursuing the goal of further developing AI systems in the direction of autonomous problem solving. Instead of simply answering individual inputs, the model is designed to plan complex tasks autonomously, coordinate intermediate steps and iteratively improve results. Gemini 3.5 is thus clearly moving in the direction of modern AI agents that not only react, but can also actively act.
According to Google, Gemini 3.5 Flash achieves performance on a par with large high-end models, but works significantly faster. The model is said to generate up to four times more output tokens per second than comparable Frontier models. This could result in considerable time advantages, especially for long workflows or complex programming tasks.
Google also highlights several benchmark results. According to the company, Gemini 3.5 Flash outperforms the previous Gemini 3.1 Pro in coding and agentic tests such as Terminal-Bench 2.1 or MCP Atlas. At the same time, multimodal understanding is improved, allowing text, images and other input to be processed more efficiently.
The focus on so-called “long-horizon tasks” seems particularly plausible. These are tasks that consist of many steps that build on each other. Examples range from software development and the maintenance of large code bases to the analysis of extensive financial data. Google is thus clearly positioning Gemini 3.5 Flash as a productivity tool for developers, companies and data-intensive work areas.
Gemini 3.5 brings new AI agents and automated workflows
A central component of Gemini 3.5 is the closer connection with Google’s new agent platform “Antigravity”. In combination with so-called subagents, the model can process and coordinate several subtasks in parallel. This makes it possible to automate extensive processes that previously required a lot of manual work.
Google provides examples of automated file management, intelligent data analysis and the creation of interactive web interfaces. Companies are also already testing the technology productively. According to Google, Shopify uses parallel AI sub-agents to evaluate large amounts of data and create more accurate growth forecasts for retailers.
In addition, Gemini 3.5 Flash will be used as a standard model in the Gemini app and in the AI mode of Google Search. Of particular interest is Gemini Spark, a new personal AI agent that will work permanently in the background and perform digital tasks independently. Its functions range from searching for information to assisting with daily work processes.
Google also claims to have made progress in terms of security. According to the company, Gemini 3.5 was developed in accordance with the Frontier Safety Framework and equipped with improved protection mechanisms against harmful content. At the same time, harmless queries should be blocked less often by mistake.
Gemini 3.5 Flash is now available worldwide – including in the Gemini app, via Google AI Studio, Android Studio and various enterprise platforms. The more powerful Gemini 3.5 Pro version is currently still undergoing internal testing and is due to be released next month.
Conclusion
With Gemini 3.5 Flash, Google is consistently driving forward the development of autonomous AI agents. In particular, the combination of high speed, multimodal understanding and autonomous task planning sets the model apart from classic chatbots. Developers, companies and productive AI applications in particular are likely to benefit from the extended agent functions. Gemini 3.5 Flash is already available worldwide, while Gemini 3.5 Pro is set to follow shortly.
