Google Introduces New Gemini Models: Focus on Speed and Security

Continuing to expand its AI ecosystem, Google has officially announced three new models developed by its DeepMind division: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. These updates focus primarily on system efficiency, processing speed, and cost reduction, opening up broad opportunities for developers and large corporations to build AI agents. This is reported by Techcrunch.com reports .

The Gemini 3.6 Flash model, the main representative of the new generation, is described by the company as a "workhorse." It has been significantly improved in coding, text processing, and multimodal tasks. Most importantly, this model has managed to reduce token usage by up to 17 percent, making it much cheaper and more economical than the previous 3.5 Flash version.

Specialized models for specific tasks

One of the highlights of the presentation was the unveiling of the Gemini 3.5 Flash Cyber model. This model is specifically trained to identify and remediate vulnerabilities in the field of cybersecurity. According to Google, this tool is currently being offered under a pilot program for a limited circle of partners and government organizations. Gemini 3.5 Flash-Lite is the most affordable model in its series, designed for projects requiring high speed with minimal costs.

According to ixbt.com, with this release, Google has focused on ensuring reliability and low latency for clients deploying AI agents at scale. However, the most anticipated news in the tech world—the update to the flagship Gemini 3.5 Pro model—did not happen this time. Gemini Pro was last updated in February, and its delay is becoming increasingly apparent in the competitive landscape.

Competition and the expected Gemini 4

While Google delays updating its most powerful models, competitors are not standing still. OpenAI has already introduced GPT-5.5 and GPT-5.6 versions, while Anthropic is strengthening its market position with powerful models like Claude Opus 4.8 and Claude Sonnet 5. According to Bloomberg, Google was forced to delay the Gemini 3.5 Pro model because it failed to meet internal performance targets.

Logan Kilpatrick, Head of Product at Google DeepMind, stated that the company is currently testing the Gemini 3.5 Pro model with partners and will release it to the public soon. He also announced that the team has begun the most massive training processes for creating the Gemini 4 model. This is expected to create a new turning point in the AI race in the near future.

This news is also of great importance for developers and startups in Uzbekistan. The lower cost and increased speed of the Flash series models will reduce the costs of creating AI services tailored for the local market, which will serve to make technological solutions more widespread.