Google DeepMind has unveiled three new additions to its Gemini family, strengthening its push to compete with rivals including OpenAI, Anthropic, and emerging Chinese AI developers. The latest releases target different use cases, ranging from general purpose AI and low cost inference to specialised cybersecurity applications.
The announcement reflects Google’s broader strategy of offering a wider range of models that balance performance, efficiency, and domain specific capabilities as demand for enterprise AI continues to accelerate.
Gemini 3.6 Flash leads the new lineup
The flagship release is Gemini 3.6 Flash, designed as Google’s primary model for developers and enterprise customers. DeepMind said the model delivers notable improvements in coding, complex reasoning, multimodal understanding, and long context tasks while reducing token usage by up to 17 percent compared with the previous generation.
Google said the model is intended to provide faster responses without sacrificing capability, making it suitable for software development, research, productivity tools, and other AI driven workflows.
Alongside stronger reasoning performance, Gemini 3.6 Flash aims to improve real world task execution, an area where Google’s Flash models have traditionally focused on balancing speed with intelligence.
Gemini 3.5 Flash Lite prioritises affordability
Google also introduced Gemini 3.5 Flash Lite, a lower cost model built for organisations running AI applications at scale.
The company said the model is optimised to deliver reliable performance while keeping inference costs to a minimum, making it well suited for high volume workloads such as chatbots, customer support, content generation, and internal business automation.
By expanding its range of lightweight models, Google is seeking to attract developers looking to deploy AI across millions of interactions without significantly increasing operating costs.
Cybersecurity model targets enterprise defence
The most specialised release is Gemini 3.5 Flash Cyber, a model fine tuned for cybersecurity tasks including vulnerability detection, validation, and software patching.
Initially, the model will be available only to governments and a limited group of trusted partners through Google’s new CodeMender pilot programme.
The launch reflects growing demand for AI powered defensive security tools as organisations increasingly look to automate threat detection and strengthen cyber resilience.
Google broadens its AI strategy
The latest Gemini releases arrive as competition across the AI sector continues to intensify. OpenAI and Anthropic continue to introduce more capable frontier models, while Chinese developers have gained attention with increasingly competitive systems.
Rather than focusing on a single flagship model, Google is expanding its portfolio to cover a broader range of customer needs, from premium enterprise deployments to cost sensitive applications and specialised industry solutions.
DeepMind has also indicated that the new models lay the foundation for Gemini 4, which the company has previously described as its most ambitious AI model to date.
The new Gemini models are rolling out through Google Vertex AI and the Gemini API, with availability varying by model and customer tier. Pricing details are expected to be announced in the coming days.
