Google has formally introduced three state-of-the-art AI models today: Gemini 3.6 Flash, Gemini 3.5 Flash Cyber, and Gemini 3.5 Flash-Lite. These innovative models strike an enhanced balance between performance and cost-efficiency. Notably, they demand 17% fewer Tokens to accomplish identical tasks when compared to their predecessors, while also offering a more budget-friendly option. Among these, Gemini 3.5 Flash Cyber stands out for its specialization in the cybersecurity domain. It excels at identifying and rectifying software vulnerabilities at a reduced cost. During testing phases, this model successfully pinpointed 10 vulnerabilities within the V8 engine that had previously eluded detection by other AI models. However, owing to security considerations, Gemini 3.5 Flash Cyber will be exclusively accessible to government entities and rigorously vetted, trusted partners. In parallel developments, Google's flagship model, Gemini 3.5 Pro, remains in the testing phase, while pre-training for the upcoming Gemini 4 has already commenced. The AI landscape is currently witnessing heightened competition, with Google encountering formidable challenges from industry players such as Anthropic, OpenAI, Moonshot AI, and Meta. Especially in the pivotal realm of cybersecurity, key competitors are also rolling out pertinent models and implementing comparable restricted access policies. Presently, Gemini 3.6 Flash and Gemini 3.5 Flash-Lite have been officially integrated into relevant applications and features. Developers can now leverage these models by accessing them through designated platforms.