Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber (blog.google)

Gemini has launched its latest AI models: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber, designed to enhance token efficiency, reduce latency, and improve overall performance for production AI agents. The 3.6 Flash model, identified as a "workhorse," decreases output token usage by 17% compared to its predecessor and delivers better coding and multimodal abilities, all at a lower cost. Alongside this, 3.5 Flash-Lite stands out as the fastest and most cost-effective in the series, boasting a throughput of 350 output tokens per second, which significantly benefits developers focusing on high-volume tasks. Additionally, the 3.5 Flash Cyber model aims to strengthen cybersecurity by efficiently identifying and addressing code vulnerabilities. It utilizes a collaborative approach within the CodeMender framework, combining multiple agents to generate comprehensive security reports. Notably, the dual-use technology is being rolled out to governments and trusted partners first, ensuring that frontline defenders can tackle security threats effectively. Overall, these advancements not only promise greater efficiency and affordability for AI workflows but also position Gemini models at the forefront of AI applications in both general and cybersecurity domains.