Skip to content
TR

What is the Gemini Decisions API? Google's yes/no AI model explained

A new Google endpoint swaps generated text for scored choices. Here is how it works, what it costs and where it fits.

By Cultvia Editorial2 min read
Tweezers placing a small black microchip onto a green circuit board
Illustrative image · Photo: Vishnu Mohanan / Unsplash

Google has added a new kind of model to its developer line-up: one that does not write answers at all. The Gemini Decisions API, announced on the Google blog on 11 November 2026, runs on a model called Gemini Decision 1 (gemini-decision-1) and is built for bounded choices such as routing a request, checking a policy rule or grading a piece of content.

What does Gemini Decision 1 actually do?

A normal chatbot call generates text token by token. Gemini Decision 1 instead takes one shared input and a list of candidate answers, then scores those options in a single pass and returns probabilities. Google describes three question types:

  • Yes/no questions, which estimate the probability that a condition holds.
  • Multiple choice, picking from between 2 and 255 options and returning a probability for each.
  • Ordered rubric scores, which grade an input on a scale and return a weighted score.

Every answer comes with a confidence score between 0 and 1, so a developer can send uncertain cases to another model or to a human reviewer. The shared input can mix text, images, audio, video and PDFs, up to 256K tokens, and one request can carry up to 1,000 independent questions.

How fast and how cheap is it?

Google's own figures put the end-to-end response at 150-250 ms for inputs of up to 16K tokens, with 50-100 ms of model compute, and say the time stays nearly flat between 1 and 100 questions per request. As Webrazzi also reported, the shared input is billed once per call, so each extra question only adds the cost of its own instructions.

Gemini Decision 1 at a glance (Google figures)
ItemFigure
Model namegemini-decision-1
Questions per requestUp to 1,000
Input sizeUp to 256K tokens
End-to-end latency150-250 ms
Input price$0.10 per 1M tokens
Output tokensNot billed
Cost vs Gemini 3.5 Flash-LiteNearly 5x cheaper on Google's benchmark suite

The benchmark and cost comparisons are Google's own measurements; no independent results have been published yet.

Where is Google already using it?

Google says Gemini Notebook uses the Decisions API as a router that sends each query to the right model tier, cutting latency by up to 40% for everyday questions, and that an experiment in its Stitch design tool uses it to apply voice or tap edits to wireframes almost instantly.

For readers: who should care, and how to try it

This is a developer tool, not a feature inside the Gemini app. Anyone building a support inbox, a content-moderation queue or an AI agent that has to choose between tools is the target user. Access is through an API key in Google AI Studio, with guides for Python, JavaScript and REST. Google has not announced a separate date or price for Türkiye, the UK or the Gulf; the blog post lists one global price. For more on how AI products are changing, browse our Reads section.

Frequently asked questions

Is Gemini Decision 1 a chatbot?

No. It does not write free text. It scores predefined answers to yes/no, multiple-choice or rubric questions and returns probabilities with a confidence score.

How much does the Gemini Decisions API cost?

Google lists $0.10 per million input tokens, with no charge for output tokens, and the shared input is billed once per request.

Cultvia Editorial

Editorial desk

Stories signed by the Cultvia editorial desk follow our editorial policy; every story lists its sources at the end and errors are fixed under our corrections policy.

#ai#Gemini#google#developers

Last updated:

What changed?
  1. : First published

Sources

  1. Gemini Decisions API: Fast, calibrated multimodal decisions — Google, 11 November 2026
  2. Google, tek API çağrısında 1.000 soruyu değerlendirebilen Gemini Decision 1'ı tanıttı — Webrazzi, 12 November 2026
Was this helpful?
Report an error

Spotted wrong information, a calculation error or a technical problem? Tell us briefly and our editors will review and fix the page.

We store the page address and browser details with your report. Your email is only used to reply, is never shared and is deleted once the report is closed.

Up next

A race track with grandstands under a cloudy blue skyTrend AnalysisRacing Bulls 2027 line-up: Lawson and Lindblad stay, Tsolov reserve13 November 2026 · 2 min read
More from Reads