Google launches a Gemini API for bounded decisions

Google introduced Gemini Decision 1 on November 11 through the Gemini API in AI Studio. It supports yes/no checks, multiple-choice selection and rubric scores across multimodal input. [1]

Spicy's takeThe company says it returns probabilities and confidence scores rather than decoding an answer token by token. That makes it an interface for bounded choices, not a substitute for every conversational task. [1]

A no-cost prototyping tier is advertised. Google says shared input is billed once per request, while each question adds its instruction and criteria tokens. [1]

Spicy's takeSpicyRanks has not tested Google's latency or accuracy claims. Evaluate uncertain cases and fallback behavior on your own workload before treating a confidence score as dependable. [1]