# Google launches Gemini Decisions API for scoring up to 1,000 questions per pass

published: 2026-11-11T21:05:31.021Z
topic: infrastructure
source: https://the-decoder.com/google-joins-the-decision-model-wave-with-gemini-decision-1-scoring-1000-questions-in-a-single-pass/
url: https://aiagentnews.io/stories/google-launches-gemini-decisions-api-for-scoring-up-to-1-000-questions-per-pass

Google DeepMind has introduced the Gemini Decisions API, built for fast closed-ended decisions such as yes/no questions, multiple choice and scale ratings. Its underlying model, Gemini Decision 1, can process up to 1,000 independent questions in one pass and is available now through Google AI Studio.

Google DeepMind has introduced the Gemini Decisions API, an interface for quick closed-ended decisions such as yes/no questions, multiple-choice picks and scale-based ratings. The underlying model, Gemini Decision 1, can handle up to 1,000 independent questions in one pass across text, images, audio, video and PDFs. It returns probabilities with confidence values in 150 to 250 milliseconds.

Google says it already uses the API internally for model routing in Gemini Notebook, design prototyping in Stitch and safety checks in Google Search. Pricing is $0.10 per million input tokens with no charge for output tokens, which Google says is about five times cheaper than Gemini 3.5 Flash-Lite. The API is available now through Google AI Studio, and integrations with OpenRouter, LangChain and Vercel are rolling out.

Google reports that Gemini Decision 1 leads its benchmark suite in agent routing, reasoning and multimodal understanding, and has the lowest Top-1 ECE calibration error. The article notes that similar models have appeared recently from the startup Jev, and that OpenAI, Cloudflare and Microsoft have followed with comparable offerings.

## Why it matters if you build agents

Developers building routers or classifiers can test whether one batched call can replace several separate model calls for closed-form decisions. Per-token pricing with no output charge could make large volumes of yes/no checks cheaper to run, though the cost comparison comes from Google. Verify the benchmark and calibration claims against Google's own materials before relying on them in production.