Gemini Decisions API: 1,000 Answers in One Go

Isometric illustration in plum purple, peach and cream showing one input splitting into many scored decisions.

Most AI models answer by writing. Google's new model answers without writing a word. The Gemini Decisions API, launched on 11 November 2026, can score up to 1,000 yes-or-no questions about a document, image or video in a single pass — in about a quarter of a second, for ten cents per million tokens.

What it is

The Decisions API is a new endpoint powered by Gemini Decision 1, built for one kind of job: bounded choices. Not open-ended text, not explanations — decisions.

You send one multimodal input — text, an image, audio, video or a PDF, up to 256,000 tokens — along with up to 1,000 independent questions. The model scores the options in a single forward pass and returns probabilities and confidence scores. There are no output tokens to pay for, because the model never generates text.

Three question types are supported: yes/no, multiple choice, and rubric scoring, where the model grades against criteria you define. Google's examples include tool selection, policy checks, model routing and context grading.

Why speed matters here

A normal language model answers one question at a time, writing tokens one after another. The Decisions API reads the input once and scores all the questions together.

Google reports end-to-end latency of 150 to 250 milliseconds, with 50 to 100 milliseconds of median model compute — nearly flat from 1 to 100 questions per request. Asking 100 questions costs about the same time as asking one.

The price matches the speed: $0.10 per million input tokens, and nothing for output. Google says it is nearly five times cheaper than Gemini 3.5 Flash-Lite with the Interactions API on its shared benchmark suite — Google's own claim, not a measured bill.

How to use it

The API is available now through the Gemini API in Google AI Studio, with a no-cost tier for prototyping. Google Cloud support is coming soon, and integrations are rolling out to OpenRouter, LangChain, Vercel and LiteLLM.

Here is how to use it, in four steps.

Step 1: Pick a decision, not a question

The API is for choices with a fixed set of answers. "Is this support ticket urgent?" works. "Write a reply to this customer" does not — that is a job for a text model. Start with the decisions your app already makes, and ask which ones are really just yes/no or multiple-choice.

Step 2: Send one input, many questions

Bundle the input once and attach every question to it. A single PDF can carry a hundred policy checks; one product photo can carry checks for damage, branding, safety and category. The input is billed once per request, however many questions you ask — batching is where the savings come from.

Step 3: Use the confidence scores

Every answer comes back with a probability and a confidence score from 0 to 1. Route the confident answers straight through, and send the uncertain ones to a bigger model or a human. Google uses the same pattern in Gemini Notebook, with up to 40% lower latency for everyday queries.

Step 4: Know the limits

At launch, the API runs on the Gemini Developer API v1beta only. Question instructions are text-only, and conversation history and tool-call blocks are not supported. It is a decision engine, not a chatbot.

Google is not first here

The idea is not new. OpenAI unveiled its own Decisions API at DevDay on 29 September 2026, Microsoft followed with Microsoft-Decision-1, built on Alibaba's Qwen, and TypeSafe launched Jev in September. Google's entry is the fastest and cheapest on its own numbers. What matters is what the category does to AI apps: every agent burns time and money asking a full language model to make small judgements, and a model that makes them in 200 milliseconds for a fraction of a cent changes the economics of the whole system.

Key takeaways

  • Google launched the Gemini Decisions API on 11 November 2026, powered by Gemini Decision 1.
  • It answers up to 1,000 yes/no, multiple-choice or rubric questions about one multimodal input in a single pass.
  • Reported latency is 150–250 ms end to end; price is $0.10 per million input tokens with no output charge.
  • Available now in Google AI Studio with a free tier for prototyping.
  • Google uses it to route queries in Gemini Notebook, reporting up to 40% lower latency.
  • It cannot generate text, remember conversation history or call tools — it only decides.

Conclusion

The Gemini Decisions API is a small, specialised tool, and that is exactly the point. Most of what AI apps do all day is not writing — it is deciding. A model that decides a thousand things in a quarter of a second, for almost nothing, quietly reshapes how other apps get built.

FAQ

What is the Gemini Decisions API?

It is a Google API, launched 11 November 2026, that scores up to 1,000 closed-ended questions — yes/no, multiple choice or rubric scoring — about one text, image, audio, video or PDF input, returning probabilities and confidence scores instead of text.

How much does Gemini Decision 1 cost?

It costs $0.10 per million input tokens, with no charge for output tokens. A no-cost tier is available in Google AI Studio for prototyping.

Can Gemini Decision 1 write text or hold a conversation?

No. It only answers fixed-choice questions. It does not generate free-form text, and it does not support conversation history or tool calls.

Sources

Related reading on Daily Tech & AI:

Researched with AI assistance and reviewed for accuracy.

Follow on WhatsApp

Follow Daily Tech & AI on WhatsApp for the latest tech and AI updates: https://www.whatsapp.com/channel/0029Vb9XzdHDJ6H4TFNAl61r

Comments