What changed
Google launched the Gemini Decisions API in Google AI Studio, powered by the new Gemini Decision 1 model. A request can score up to 1,000 yes/no, choice or rubric questions across text, images, audio, video and PDFs, returning probabilities and confidence rather than generated prose.
Why it matters
Purpose-built scoring can move tool selection, policy checks and routing out of expensive generative calls. The practical value is calibrated, auditable choices at stated low latency and price, not another general chatbot.
What to do
Run a shadow evaluation on one closed-set routing or policy task, compare calibration and false-positive costs with the current classifier, and keep a human or deterministic rule on irreversible decisions.
Caveat
Google reports 150–250 ms end-to-end latency, $0.10 per million input tokens and benchmark leadership. Those are first-party results, and the flat-latency claim applies to tested contexts and question counts rather than every production workload.