Google DeepMind Unleashes Faster, Cheaper AI Models

Google DeepMind has released a suite of new Gemini AI models: 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. The 3.6 Flash model is designed as a versatile tool, enhancing coding, knowledge tasks, and multimodal performance while reducing token usage by up to 17% compared to its predecessor, 3.5 Flash, making it more cost-effective.

The 3.5 Flash-Lite model stands out as the most budget-friendly option, whereas the 3.5 Flash Cyber is a specialized offering fine-tuned for identifying and rectifying cybersecurity weaknesses efficiently. Google plans to offer this Cyber model exclusively to governments and trusted collaborators through a limited pilot program.

These releases underscore Google's commitment to delivering efficiency, low latency, and reliability for large-scale AI agent development. However, the company has not yet released an update to its premier Gemini Pro model, which was last updated in February.

This delay occurs amidst fierce competition, with rivals OpenAI and Anthropic continually launching new iterations of their AI systems. Google DeepMind product lead Logan Kilpatrick indicated that Gemini 3.5 Pro is currently undergoing partner testing and is expected to be released soon. The team has also commenced its most extensive pre-training initiative to date for Gemini 4.