Google Unveils Three New Gemini AI Models, Expands Focus on Speed, Efficiency and Cybersecurity: MOUNTAIN VIEW, Calif. — Google has expanded its Gemini artificial intelligence portfolio with the launch of three new AI models, introducing Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. The company says the latest additions are designed to deliver faster performance, improved efficiency, and specialized capabilities, while offering developers more choices for different AI workloads.
Alongside the product launches, Google also shared an update on its upcoming flagship model, Gemini 3.5 Pro, confirming that it remains in partner testing. The company further revealed that pretraining for the next-generation Gemini 4 model has already begun, signaling its continued investment in advancing generative AI.
The announcements come as competition among leading AI companies—including Google, OpenAI, Anthropic, and Meta—continues to accelerate. While flagship AI models often dominate headlines, Google’s latest releases focus on improving practical applications such as software development, enterprise productivity, agentic workflows, and cybersecurity.
Gemini 3.6 Flash Becomes Google’s New Workhorse AI Model
At the center of Google’s latest AI lineup is Gemini 3.6 Flash, which the company describes as its new “workhorse” model for developers and businesses.
According to Google, the model delivers stronger coding capabilities, better reasoning for knowledge-intensive tasks, and enhanced multimodal understanding, enabling it to process text, images, and other data more effectively.
One of the biggest improvements is greater token efficiency. Google claims Gemini 3.6 Flash can reduce token usage by up to 17% compared to its predecessor while also lowering the cost per processed token. Reduced token consumption allows developers to complete AI tasks using fewer computational resources, helping lower operating costs without sacrificing performance.
The company says these improvements were driven by extensive feedback from developers and enterprise customers, who requested faster response times and better efficiency for production applications.
Internal benchmark results released by Google also show improvements in coding accuracy, reasoning ability, and average tokens consumed per task.
Gemini 3.5 Flash-Lite Prioritizes Speed and Cost Efficiency
Google also introduced Gemini 3.5 Flash-Lite, positioning it as its fastest and most affordable AI model.
The company says Flash-Lite is capable of generating up to 350 output tokens per second, making it particularly well suited for applications requiring real-time responses such as chatbots, customer support, search experiences, and lightweight AI assistants.
Despite its focus on speed, Google says Flash-Lite significantly outperforms previous generations in handling agentic workflows—tasks where AI systems independently plan, execute, and complete multi-step objectives.
A notable enhancement is built-in computer-use capability, allowing the model to interact with digital interfaces and software applications as part of larger automated workflows.
Google believes this functionality will help developers build more capable AI agents that can complete tasks with minimal human intervention.
Cybersecurity Gets Dedicated AI Model
Perhaps the most specialized announcement is Gemini 3.5 Flash Cyber, a new AI model developed specifically for cybersecurity operations.
Unlike general-purpose AI systems, Flash Cyber has been optimized to identify software vulnerabilities, assist with security investigations, and accelerate the process of fixing bugs.
The model works alongside Google’s infrastructure agent, CodeMender, which helps security teams analyze codebases, locate vulnerabilities, recommend fixes, and automate portions of the software patching process.
According to Google DeepMind, the cybersecurity-focused AI is already being used internally across several major Google projects, including Android, Chrome, and YouTube, where it has successfully identified and resolved software issues before they reached production environments.
Initially, access to Gemini 3.5 Flash Cyber will be limited to government agencies and trusted partners, reflecting the sensitive nature of cybersecurity tools. Google says broader availability is planned in the future.
Gemini 3.5 Pro Remains in Testing
While the newly announced Flash models are available immediately, Google’s most anticipated AI system—Gemini 3.5 Pro—is still under development.
The company confirmed that the flagship model is currently undergoing testing with select partners. Although no official launch date has been announced, Google says it intends to release Gemini 3.5 Pro once it meets its quality and reliability standards.
The update suggests that Google is prioritizing performance and stability before making its highest-end model broadly available.
Google Already Looking Toward Gemini 4
Even as Gemini 3.5 Pro approaches release, Google revealed that work has already begun on its next major AI platform.
The company confirmed that pretraining for Gemini 4 is underway, marking the earliest stage of development for what is expected to become Google’s next flagship family of AI models.
Although Google did not provide technical details or a release timeline, the announcement highlights the rapid pace of AI innovation, with companies investing in future generations even before current flagship models become publicly available.
Growing Competition in the AI Industry
Google’s latest announcements arrive amid an increasingly competitive AI landscape.
Major technology companies continue releasing improved language models at a rapid pace, competing across coding, reasoning, multimodal understanding, enterprise productivity, and autonomous AI agents.
Rather than focusing solely on building larger models, companies are increasingly optimizing AI for specific use cases. Google’s latest releases reflect this strategy by offering separate models tailored for high-performance workloads, low-cost deployment, and cybersecurity applications.
Industry analysts say this diversification gives developers greater flexibility when choosing AI models based on their performance requirements and budget constraints.
Chinese Scientists Develop Low-Cost Technology to Turn Plastic Waste into Jet Fuel | Maya
Availability
Google announced that Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are available immediately for developers through the Gemini API, Google AI Studio, Android Studio, and the Gemini app.
In addition, Gemini 3.6 Flash has been integrated into Google Antigravity, while Gemini 3.5 Flash-Lite is rolling out across Google Search, bringing faster and more efficient AI capabilities to millions of users.
The company expects the new models to support a broad range of applications, from software development and customer service to enterprise automation and cybersecurity, as Google continues expanding its AI ecosystem.
