Google Brings Key Updates to AI Agents, Announces Gemini 2.5 Flash
At the Google Cloud Next 2025 event, Google unveiled significant updates related to AI agents. These updates include new tools, protocols, and frameworks that empower developers and enterprises to efficiently build, connect, and deploy agentic systems within their ecosystems.
Gemini 2.5 Family Expansion
During the event, Google introduced a new addition to the Gemini 2.5 family of models—the Gemini 2.5 Flash. This new model aims to enhance AI capabilities and performance for users.
Agent2Agent (A2A) Protocol
Google also announced the launch of the Agent2Agent (A2A) protocol, designed to facilitate seamless communication between AI agents operating within an enterprise ecosystem. This open protocol enables AI agents to interact with each other regardless of the framework or vendor they are built on.
Agent Development Kit (ADK)
As part of the updates, Google introduced the Agent Development Kit (ADK), an open-source framework that enables the design of AI agents and multi-agent systems with enhanced control and customization options. The ADK currently supports Python with plans to expand language support in the future.
Agent Engine and Agentspace
Google also announced the Agent Engine, a platform that simplifies the deployment of AI agents across different frameworks and models. This deployment tool handles various aspects such as agent context, infrastructure management, scaling complexities, security, evaluation, and monitoring.
Gemini 2.5 Flash Model
Google introduced the Gemini 2.5 Flash model, available on Google's Vertex AI and Google AI Studio platforms. Positioned as a "workhorse model", the Gemini 2.5 Flash prioritizes low latency and cost efficiency while offering users control over reasoning time based on query complexity.
Future Developments
Looking ahead, Google plans to integrate Gemini models into the Google Distributed Cloud (GDC) in collaboration with NVIDIA, offering organizations the flexibility to run applications closer to their data sources. The company also unveiled Ironwood, the seventh generation tensor processing unit (TPU), known for its enhanced performance and efficiency compared to its predecessors.
Stay tuned for more updates and advancements in AI technology from Google as they continue to push the boundaries of innovation and connectivity.
