Home / Technology / Artificial Intelligence
TECHNOLOGY · ARTIFICIAL INTELLIGENCE · GOOGLEGoogle’s New Gemini Model Can Make 1,000 Decisions in One Pass — Here’s What It Actually Does
Google has introduced Gemini Decision 1, a model designed to score up to 1,000 bounded questions in one request. Rather than writing a conversational answer, it returns probabilities and confidence scores for choices such as yes/no checks, multiple-choice decisions and rubric-based ratings.
A Gemini model built to choose, not chat
Google’s newest Gemini release is aimed at a different job from a typical chatbot. Announced on November 11, Gemini Decision 1 powers the company’s new Decisions API, which is designed for tasks where software needs to select or score a defined set of options rather than produce a paragraph of text.
Google says a single request can contain up to 1,000 questions against shared input. The model supports text, images, audio, video and PDFs, and can return probabilities and a confidence score for each decision. Google lists examples such as checking whether a policy condition is met, choosing a tool for an AI agent, routing a request to the right model or grading content against a rubric.
How the 1,000-question approach works
Instead of generating words one token at a time, Gemini Decision 1 evaluates a bounded question against candidate answers. Google describes three decision types: yes-or-no, multiple choice and ordered rubric scoring. A developer can reuse the same input across many questions, which may be useful when a system needs to classify or evaluate a batch of items.
For example, a business could provide a support conversation and ask several separate questions about whether a refund policy applies, which department should handle the case and how urgent the request is. A document-processing system could examine a form and related images, then score multiple checks in one request. These are illustrative use cases, not claims that the model makes every such decision correctly.
Speed and price: what Google says
Google says the API has end-to-end latency of about 150 to 250 milliseconds and costs $0.10 per million input tokens, with no output tokens billed. The company also says the shared input is billed once per request, regardless of how many questions are asked, while each question adds its own instruction and criteria tokens. Google says a no-cost tier is available for prototyping in Google AI Studio.
Those figures are Google's published product claims and pricing details. Actual response times and total costs can vary with request size, input type, question instructions, usage limits and production requirements. Developers should check the current documentation before building a cost estimate.
What it does not mean
The release does not mean Gemini has become a general-purpose reasoning system that can safely make any decision without oversight. The API is designed for closed-set choices: developers must define the questions and the available answers or scoring rubric. Confidence scores can help route uncertain cases to another model or a human reviewer, but they are not a guarantee that an answer is correct.
For consequential decisions involving people, money, employment, healthcare or legal rights, developers still need testing, monitoring, clear escalation paths and human review. A fast classification is only useful when the categories are well designed and errors are handled responsibly.
Why this matters for AI agents and software teams
AI applications often spend time asking a large language model to perform small checks: classify a request, choose a tool, rank candidates or decide whether an input meets a rule. A model built specifically for those bounded tasks could reduce latency and cost in some workflows. It may also make systems easier to evaluate because the output is a score or selected option rather than a long generated response.
The broader trend is a move towards specialised AI components. Instead of using one large model for every step, developers can combine generative models with smaller decision systems, databases and conventional code. Whether that architecture performs better will depend on the task and the quality of the evaluation process.
Availability
Google says Gemini Decision 1 is available through the Gemini API in Google AI Studio. The company also lists integrations rolling out across developer tools and frameworks. Availability, limits and supported SDK versions may change, so developers should consult Google's live documentation before relying on a specific integration.
Source
Editorial note: Product capabilities, latency and pricing in this article are attributed to Google's announcement and have not been independently benchmarked by Rajasthan News 24. Contact the newsroom to report a correction.