Google releases three new Gemini models and Gemini Robotics ER 2 for agentic workflows

Google’s July 2026 recap post highlights a wave of AI advances, including three new Gemini models tailored for developers building production-grade agentic systems. The company describes these models—Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber—as optimized for higher token efficiency, lower latency, and more reliable performance to enable scaling of agent workflows.

The post also introduces Gemini Robotics ER 2, positioned as the company’s most capable embodied reasoning model to date. Google frames ER 2 as a bridge between digital reasoning and physical action, enabling systems to engage in natural conversation, interpret surroundings, and execute multi-step tasks in real-world environments.

Beyond models, the recap includes updates to creative tooling, such as next-generation music and video features, and tighter integration between Search and Gemini Spark for managing daily tasks. Google also highlights new capabilities for transferring data between devices, including a native migration experience for iPhone users switching to Android 17.

The company underscores its broader efforts to apply AI to societal challenges, including early wildfire detection via new satellites and a partnership to support the skilled trades workforce. Google DeepMind CEO Demis Hassabis is quoted reflecting on the company’s vision for AI’s role at this stage of development.