Google's new Gemini Decisions API answers up to 1,000 closed questions in a single call

On November 11, Google introduced the Gemini Decisions API, a new developer endpoint built for closed decisions: yes or no, multiple choice and rubric-based scores. The tool runs on the new Gemini Decision 1 model and is available through the Gemini API in Google AI Studio.

According to the announcement published the same day on The Keyword, Google's blog, the model takes a single input, which can mix text, images, audio, video and PDFs up to 256K tokens, along with up to 1,000 independent questions. Instead of writing out an answer word by word, it scores the possible options in a single pass and returns probabilities and confidence scores.

Google reports end-to-end latency of 150 to 250 ms and pricing of $0.10 per 1 million input tokens, with no charge for output tokens. The shared input is billed once per request, no matter how many questions it carries. By the company's own comparison, the tool comes in almost 5 times cheaper than Gemini 3.5 Flash-Lite with the Interactions API on the same set of tests.

Three kinds of decisions

Each question can be one of three types: yes or no, multiple choice with 2 to 255 options, or a score on an ordered scale that starts at 0. Every answer comes with a confidence score from 0 to 1, so a system can route ambiguous cases to a fallback or to a human.

On the agent routing, reasoning, ranking and multimodal understanding tests the company cites, Gemini Decision 1 leads the field and posts the lowest calibration error. There is also a free tier in Google AI Studio for prototyping.

How Google uses it internally

The company says internal teams already rely on the API. Gemini Notebook uses it to send each question to the right model tier, cutting latency by up to 40% on everyday queries. CC, from Google Labs, picks an emoji reaction from more than 200 options, and Stitch generates clickable screen drafts in seconds.

In Google Search, the API serves as a safety layer for MCP tools, detecting adversarial behavior and content within the service's latency budget.

Integrations are rolling out to OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode, as well as the google-genai SDKs. Gemini support on Google Cloud is coming soon.

For brand teams, the takeaway is that Google now offers a low-cost way to classify content and check policy rules across text, images, audio and video in a single call, and that Search itself already uses this kind of fast decision to catch adversarial content.