Google Launches Agentic Vision in Gemini 3 Flash

Summary

Google DeepMind launched Agentic Vision for Gemini 3 Flash, a capability that transforms image understanding from a static single-pass process into an active, multi-step investigation. Rather than analyzing an image in one shot, the model operates through a "Think, Act, Observe" loop: it formulates a multi-step plan, executes Python code to zoom in, crop, annotate, and manipulate images, then appends the transformed output back into its context window to support a more grounded final answer. Ke