Gemini Robotics-ER 1.6 Preview is documented as part of Google's Gemini API lineup, with its dedicated developer page published on the official Gemini API documentation site, marking it as a preview entry intended for robotics-oriented embodied reasoning workloads. The preview designation points to an experimental release that invites early integration feedback before a general-availability iteration, and it sits alongside the broader Gemini family of multimodal models that share a common design lineage for fusing language and perception signals into a single interface.

From a practical standpoint, the preview accepts rich multimodal input in the form of text, images, video, and audio, while returning text-only outputs, which makes it well suited to pipelines that translate perception into instructions, plans, or structured decisions for downstream robotic control. The combination of a 131,072-token context window and a 65,536-token maximum output gives the model room to ingest long scene descriptions, trajectories, or interaction histories and to produce detailed step-by-step reasoning in return, while listed support for reasoning, tool calling, and temperature control lets developers tune behavior and connect the model to external planners or action modules.