Featured:MeloLab– Create songs and background music with AI.

Gemini Decisions API

Google's Gemini Decisions API scores candidate options in one pass, returning probabilities and confidence scores in 150–250 ms at $0.10 per 1M input tokens.

Type:
AI Tool
Gemini Decisions API product preview
Contents

What is Gemini Decisions API?

Google's Gemini Decisions API scores candidate options in one pass, returning probabilities and confidence scores in 150–250 ms at $0.10 per 1M input tokens.

Updated ·

Gemini Decisions API: Fast Closed-Set Decisions with Gemini Decision 1

Announcement date:

The Gemini Decisions API is a new endpoint from Google for fast, closed-set decisions powered by the Gemini Decision 1 model (gemini-decision-1). Instead of generating answer text token by token, it scores candidate options in a single forward pass and returns probabilities and confidence scores. It is available through the Gemini API in Google AI Studio, with a no-cost tier for prototyping.

Social artwork for the Gemini Decisions API showing the title 'Fast, calibrated multimodal decisions'.
Publisher announcement artwork; not a product interface screenshot.

What the Gemini Decisions API is

The Gemini Decisions API is a dedicated endpoint for making closed-set decisions rather than generating free-form text. It is powered by the new Gemini Decision 1 model (gemini-decision-1) and is available through the Gemini API in Google AI Studio.

The key difference from text generation is the output: Gemini Decision 1 does not produce answer text token by token. Instead, it scores your candidate options in a single forward pass and returns probabilities and confidence scores. This makes it suited to classification, routing, and grading tasks where the possible answers are already known.

  • Closed-set decisions: you supply the candidate options, the model scores them.
  • Returns probabilities and confidence scores rather than generated text.
  • Available through the Gemini API in Google AI Studio.

Supported decision types: noul, choice, and score

The API supports three decision types. noul is a yes/no decision that estimates the probability that a yes/no condition holds, with optional true and false description criteria. choice is multiple choice: it selects from 2 to 255 choices and returns the winner plus normalized probabilities for every candidate. score is ordered rubric grading: it grades inputs across an ordered rubric starting at 0 and returns a continuous weighted score and per-level probabilities.

  • noul: yes/no probability with optional true and false criteria.
  • choice: 2 to 255 options, returns the winner and normalized probabilities for all candidates.
  • score: ordered rubric starting at 0, returns a continuous weighted score and per-level probabilities.

Multimodal input and many questions per request

A single request can include one multimodal input spanning text, images, audio, video, and PDFs up to 256K tokens, alongside up to 1,000 independent questions. This lets you evaluate many decisions against the same shared context without resending it.

Billing reflects that sharing: the shared input is billed once per request regardless of how many questions you evaluate, and each question adds only its own instruction and criteria tokens.

  • One multimodal input across text, images, audio, video, and PDFs up to 256K tokens.
  • Up to 1,000 independent questions per request.
  • Shared input billed once per request; each question adds only its instruction and criteria tokens.

Pricing and latency

Gemini Decision 1 is priced at $0.10 per 1M input tokens with zero output tokens billed, and it offers end-to-end latency of 150–250 ms. A no-cost API tier is available in Google AI Studio for prototyping.

  • $0.10 per 1M input tokens; output tokens are not billed.
  • End-to-end latency of 150–250 ms.
  • No-cost API tier in Google AI Studio for prototyping.

Confidence scores and human routing

Every answer includes a confidence score from 0 to 1. This lets your code route ambiguous cases to a fallback or a human reviewer instead of treating every decision as equally reliable. A practical approach is to set a threshold appropriate to your use case and send low-confidence results to review.

  • Confidence score from 0 to 1 on every answer.
  • Use thresholds to route ambiguous cases to a fallback or human reviewer.
  • Threshold values are a design choice, not a documented default.

SDKs, integrations, and getting started

The API is available through the Gemini API in Google AI Studio, where a no-cost tier supports prototyping. Ecosystem integrations are rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit, and OpenCode alongside the google-genai SDKs, with support coming soon to Gemini in Google Cloud.

To get started, begin in Google AI Studio with the no-cost tier, define your candidate options and criteria, and test how confidence scores behave on your own inputs before relying on thresholds in production.

  • Start in Google AI Studio with the no-cost tier.
  • google-genai SDKs are available; integrations are rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit, and OpenCode.
  • Support is coming soon to Gemini in Google Cloud.

Example use case: Gemini Notebook

Gemini Notebook uses the Decisions API as a real-time router to classify intent in milliseconds and serve everyday queries with up to 40% lower latency, while reserving deep reasoning models for complex tasks. This illustrates the intended pattern: fast closed-set routing in front of heavier generation.

  • Real-time intent classification in milliseconds.
  • Up to 40% lower latency for everyday queries.
  • Deep reasoning models reserved for complex tasks.

Questions

How is Gemini Decision 1 different from text generation?

Gemini Decision 1 does not generate answer text token by token. It scores your candidate options in a single forward pass and returns probabilities and confidence scores, which suits closed-set decisions such as classification, routing, and rubric grading.

What decision types does the Gemini Decisions API support?

It supports three types: noul (yes/no probability with optional true and false criteria), choice (2 to 255 options, returning the winner and normalized probabilities for every candidate), and score (ordered rubric starting at 0, returning a continuous weighted score and per-level probabilities).

How much does Gemini Decision 1 cost and how fast is it?

Gemini Decision 1 is priced at $0.10 per 1M input tokens with zero output tokens billed, and offers end-to-end latency of 150–250 ms. A no-cost API tier is available in Google AI Studio for prototyping.

Original sources

  1. blog.google — Original source 1 ↗

Tool Info

Has this tool changed? Help us keep it up to date

See also: related tools and guides

Loading reviews…

Gemini Decisions API Reviews

Not rated yet0 user reviews

Do you recommend Gemini Decisions API? Leave a comment below!

Your rating
0 / 2000

Loading reviews…

Categories

FOR PRODUCT BUILDERS

Built something useful? Help people find it.

Start with your website. Add your logo, screenshots and product details to a listing you can keep up to date.

Verify your website badge for a free listing.