Google ships the Gemini Decisions API, no output tokens billed
A new endpoint scores up to 1,000 yes/no, multiple-choice or rubric questions in one pass across text, images, audio, video and PDFs, on a model built to never generate text.
Google has launched the Gemini Decisions API, built on a new model called Gemini Decision 1 (gemini-decision-1), for anything that needs a bounded answer rather than generated text. One call can carry up to 1,000 yes/no, multiple-choice or rubric-scored questions, across text, images, audio, video and PDF input up to a 256,000-token context, and returns in 150 to 250 milliseconds.
result = client.decisions.create(
model="gemini-decision-1",
input=[{"type": "text", "text": "I was charged twice."}],
questions={"refund": {"type": "noul", "instructions": "Is a refund being requested?"}},
)
Pricing is $0.10 per million input tokens, with nothing charged for output, and the shared input across a batch is billed once no matter how many questions you ask of it. Google says the model beats Gemini 3.5 Flash-Lite on calibration error at roughly a fifth of the cost, and it is live now through Google AI Studio and the Gemini API, with OpenRouter, LangChain and Vercel integrations following.
Why it matters: OpenAI opened its own Decisions API five weeks ago, and this is Google’s answer. For routing, moderation and classification calls, the ones that only ever need a label back, a full chat completion was always the wrong tool. Two labs now price the right one cheaply enough that there is no excuse left to keep using the wrong one.