Google DeepMind Introduces Gemini 3.6 Flash, 3.5 Flash Lite, and 3.5 Flash Cyber

Google's DeepMind team has announced three new additions to its Gemini model family: Gemini 3.6 Flash, 3.5 Flash Lite, and 3.5 Flash Cyber. The models aim to improve efficiency and reliability for developers building large scale AI agents.

Gemini 3.6 Flash is described as the primary model in this release, offering better coding and multimodal performance while cutting token usage by up to 17 percent compared to the previous 3.5 Flash version. It operates at lower cost and includes enhanced safeguards against misuse in sensitive domains.

Gemini 3.5 Flash Lite focuses on low latency and high throughput workloads, producing up to 350 output tokens per second. It is the most cost-effective model in the 3.5 class and is designed for tasks such as search and document processing.

Gemini 3.5 Flash Cyber is a security-focused model created for identifying and resolving software vulnerabilities. It is being released through a limited access pilot for governments and trusted partners.

The company also confirmed that testing continues for Gemini 3.5 Pro, which has not yet been publicly released. Work has already begun on Gemini 4, described as DeepMind's most ambitious training run to date.