Google expanded its artificial intelligence lineup on Tuesday, unveiling three new models built on the Gemini Flash architecture. While these updates do not include the highly anticipated flagship release of Gemini 3.5 Pro, they represent a significant step forward in efficiency and speed. In a market where companies like OpenAI and Anthropic are releasing new tools almost weekly, Google is focusing on practical, specialized utility over raw power for this particular rollout.
The centerpiece of the announcement is Gemini 3.6 Flash, described by the company as its workhorse model. Designed to be more reliable and cost-effective than its predecessor, it boasts a seventeen percent reduction in token usage while improving performance in coding and multimodal tasks. Alongside it comes Gemini 3.5 Flash-Lite, the leanest version of the family designed for extreme speed. Capable of delivering three hundred fifty output tokens per second, Lite is specifically tuned for agentic workflows and includes built-in tools for computer use.
Perhaps the most niche addition is Gemini 3.5 Flash Cyber, a specialty model dedicated entirely to cybersecurity. Working in tandem with an infrastructure agent known as CodeMender, this model helps security teams hunt down and fix software vulnerabilities. Google reports that the system is already being used internally to scrub bugs from Android, Chrome, and YouTube before it eventually expands beyond government agencies and trusted partners.
Despite these additions, many users remain focused on when Gemini 3.5 Pro will finally arrive. Google confirmed that the flagship model is currently undergoing partner testing and will launch once it reaches readiness. To keep the momentum going, the company revealed it has already started pretraining for Gemini 4. For now, developers can access 3.6 Flash and 3.5 Flash-Lite through various APIs and studios, with Lite beginning its integration into Google Search.
