Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

Google has unveiled three new AI models as part of the Gemini series: 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. The Flash series models are designed to provide higher token efficiency, lower latency, and more reliable performance for AI agents. Building upon the previously announced Gemini 3.5 Flash, the new models aim to strike a balance between efficiency and quality to support scaling agentic workflows.

Gemini 3.6 Flash, the latest in the Flash series, incorporates feedback from developers and customers, resulting in a 17% reduction in output tokens compared to 3.5 Flash. This enhanced efficiency is accompanied by fewer reasoning steps and tool calls, leading to cost savings and improved performance in complex workflows and knowledge-based tasks. The model is also more resistant to jailbreaks and has been fine-tuned for cybersecurity vulnerability detection and patching.

In contrast, 3.5 Flash-Lite is tailored for low-latency tasks and high throughput scenarios, such as agentic search and document processing. With a remarkable 350 output tokens per second, it boasts a strong price-to-performance ratio compared to its predecessor, 3.1 Flash-Lite. This model excels in handling high-volume tasks, providing efficient scaling for agentic systems.

Lastly, 3.5 Flash Cyber is a specialized version of 3.5 Flash, fine-tuned specifically for cybersecurity vulnerability detection and resolution. It operates at a lower price point than larger models and is currently available through a limited-access pilot program to governments and trusted partners. This targeted release aims to bolster frontline cybersecurity defenses while minimizing the risk of broader misuse.