Event date · · Google DeepMind
Gemini 2.0 Launch: Google Advances Multimodal Models to Agent Systems
Google releases Gemini 2.0 Flash and showcases prototypes including Project Astra, Mariner, and code agents.
Model capabilities are organized into a complete system of perception, reasoning, tool use, and action. Google begins integrating search, browser, Android with the model.
Native multimodal input/output, tool calling, and low-latency inference make real-time agents the main product line.
Platforms with browser, mobile OS, and search entry points gain structural distribution advantages.
Independent agent products must possess industry processes, data, or end-to-end delivery capabilities that Google cannot easily replicate.
Observe the speed of prototype-to-product transition, as well as cross-product permissions and user trust.