Google enters a new phase in the AI race: Gemini 3.6 and Gemini 4 announced

Tech giant Google continues to expand its AI ecosystem. The company has officially introduced three new models in the Gemini family: Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and the specialized Gemini 3.5 Flash Cyber. These updates aim not only to increase computational speed but also to make AI integration into daily workflows more affordable and efficient. This is reported by Ixbt.com reports .

The flagship of the new line, the Gemini 3.6 Flash model, is significantly more cost-effective than the previous generation, consuming 17% fewer output tokens. According to Artificial Analysis Index data, the increase in efficiency has not led to a decrease in quality; on the contrary, performance in programming, data analysis, and multimodal content processing has improved. Google has also lowered the price of this model, setting it at $1.50 per 1 million input tokens and $7.50 for output tokens.

A revolution in speed and cybersecurity

The second innovation is the Gemini 3.5 Flash-Lite model, specifically designed for high speed and handling large volumes of requests. Capable of generating up to 350 tokens per second, this model has become the fastest system in its class. Despite a slight price increase, it has proven to be far superior to its predecessors in complex agentic scenarios and working with long-text contexts.

The company has also paid special attention to the field of cybersecurity by introducing the Gemini 3.5 Flash Cyber model. This model is optimized for finding and remediating vulnerabilities in software code. It works in collaboration with the CodeMender agent and demonstrates high efficiency in preparing complex security reports. For now, this model is only available under a limited pilot program for government organizations and trusted partners.

Test results presented by Google clearly demonstrate the potential of the new models. For example, Gemini 3.6 Flash recorded a 49% result on the DeepSWE benchmark, which evaluates programming capabilities, whereas the previous version showed 37%. Additionally, computer control capability reached 83% in the OSWorld-Verified test, indicating that AI's ability to interact with operating systems has expanded.

A step toward the future: Gemini 4 is in development

Alongside the release of the new models, Google announced that it has begun work on a larger project: the Gemini 4 generation. According to company representatives, the most extensive phase of pre-training for the next generation of models is currently underway. This confirms Google's intention not to lose its lead in the competition against OpenAI and other rivals.

For developers and tech startups in Uzbekistan, these updates are significant, as the Gemini 3.6 Flash and Gemini 3.5 Flash-Lite models are already open for use on the Gemini API, Google AI Studio, and Android Studio platforms. This enables the creation of fast and affordable AI-based applications in the local market.