Google Boosts AI Race with High-Speed, Security-Focused Gemini Models

Google has intensified its competitiveness in the artificial intelligence market with the launch of a new family of Gemini models. The updates seek to improve speed, efficiency and security, aiming to consolidate the company’s position in a scenario of intense technological competition, especially against rivals such as OpenAI and Anthropic. The initiative reflects a strategic bet to expand the company’s presence in critical areas of AI, such as intelligent agents and cyber protection.
Innovations come to Gemini models with a focus on performance
New features include Gemini 3.6 Flash, considered the most advanced Flash model developed by the company. It offers better programming performance, diverse knowledge tasks, and multimodal capabilities, while consuming 17% fewer tokens than its previous version. This optimization results in a significant reduction in usage costs for developers and companies. The bet is clear: offering a balance between quality and accessibility to expand the adoption of technology in various applications.
Tulsee Doshi, senior director of product management on the Gemini team, said, “The new Flash models are designed to hit the sweet spot between efficiency and quality.” This statement underlines Google’s intention to balance computing power with practicality and cost-benefit.
Specialization strengthens defense against cyber threats
Digital security gained a dedicated model: the Gemini 3.5 Flash Cyber. This model was specifically tuned to identify and fix software vulnerabilities, offering an extra layer of protection. Google reinforces that this version has received significant improvements to resist attempts to misuse it in cyber attacks, without compromising legitimate applications.
In internal tests released by the company, Gemini 3.5 Flash Cyber managed to identify 55 problems in the V8 JavaScript Engine. Of these, 10 vulnerabilities had not been detected by other AI models. Due to its critical nature, the initial availability of this model will be restricted to governments and partners considered trustworthy, highlighting the seriousness of the security application.
Lighter versions and autonomous agents drive scale
To meet specific demands, Google also introduced Gemini 3.5 Flash-Lite, which was created for high-volume tasks and the development of AI agents capable of performing actions autonomously. This version seeks to optimize performance in scenarios where scale and automation are essential.
- Gemini 3.6 Flash, designed to be more efficient and cost-effective;
- Gemini 3.5 Flash Cyber, aimed specifically at digital security;
- Gemini 3.5 Flash-Lite, ideal for AI agents and large-scale operations;
- Start of pre-training for Gemini 4, the next generation of the platform.
The arrival of Gemini 3.6 Flash and Flash-Lite marks the beginning of the release for developers, companies and users of the Gemini application. Flash-Lite, in particular, will be gradually integrated into Google Search, expanding its reach and impact.
Next generation and continuous testing define the future of Gemini
While the new models are being implemented, Google reported that the Gemini 3.5 Pro remains in the intensive testing phase and will only be released to the public when it is completely ready. This cautious approach aims to ensure robustness and quality before broad availability.
At the same time, the company has already started pre-training Gemini 4, signaling that the next generation of the platform is already in advanced development. This constant movement of innovation reflects the dynamic nature of the race for artificial intelligence. Companies like Google, OpenAI, Anthropic and Meta are in fierce competition, seeking increasingly efficient and specialized models, where cost, performance and security have become factors as crucial as the quality of the responses generated.






