Google has launched the Gemini Decision 1 model and its accompanying Gemini Decisions API, designed specifically for decision-making tasks with a closed set of options. Unlike traditional generative large models that produce answers token by token, this model scores preset candidate answers through a single forward pass. It supports three core capabilities: yes/no judgments, multiple choice with up to 255 options, and ordered scoring based on custom criteria—and all three task types can be executed together within a single request. The model supports multimodal input of up to 256K tokens, can process up to 1,000 independent questions in a single request, and has an end-to-end latency of only 150–250 milliseconds. It is priced at $0.1 per million input tokens, with no charge for output tokens—about one-fifth the cost of Gemini 3.5 Flash-Lite. In relevant benchmark evaluations, the model ranked first in 7 of 9 tests, and its probability calibration error was far lower than that of multiple competitors. Google has already applied it to scenarios such as model routing, Agent tool selection, product design, and security detection. Internal implementation cases show that it can reduce Agent routing latency by up to 40%.