The Gemini Decisions API is a Google endpoint that answers closed-set questions (yes/no, multiple choice and rubric scores) by scoring the options in one pass instead of writing an answer. Google launched it on November 11, 2026. It's built for the many small decisions inside AI agents, such as which tool to call or whether a reply breaks a policy, and it charges nothing for output.
Gemini Decisions API at a glance
Price | $0.10 per 1M input tokens, $0 output. No-cost tier in Google AI Studio. |
|---|---|
Model | gemini-decision-1 |
Speed | 150-250 ms end to end (Google's figures) |
Inputs | Text, images, audio, video and PDFs, up to 256K tokens |
Questions per request | Up to 1,000 |
Where | Gemini API and Google AI Studio (Google Cloud support "coming soon") |
How it works
You send an input (a conversation, a document, an image) plus a list of questions. Each question is one of three types: yes/no, multiple choice with 2 to 255 options, or an ordered score against a rubric. The model scores every candidate answer in a single forward pass and returns each answer with a confidence score from 0 to 1. Because the shared input is billed once per request, asking 50 questions about the same document costs barely more than asking one.
What you can do with it
Route each user request to the right model or tool before the main model runs.
Check replies against content or brand policies at scale.
Grade whether retrieved documents are relevant before they go into a prompt.
Score support tickets, leads or reviews against a rubric.
Pricing
Input costs $0.10 per million tokens and output is free. Google says it works out nearly 5x cheaper than Gemini 3.5 Flash-Lite across its shared benchmark suite. Google's Gemini Notebook team reports up to 40% lower latency on everyday queries after using it for routing.
What to know
It only answers questions with a fixed set of answers. It doesn't write text, so it complements a generative model rather than replacing one.
Speed, cost and accuracy figures are Google's own.
Integrations are rolling out with OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode.
Our take
This is plumbing, but it's plumbing every agent builder overpays for. Most teams answer "is this allowed?" or "which tool next?" by asking a full chatbot to write the word "yes." If the latency holds up in your tests, you can put dozens of these checks in front of every agent step without users noticing.
Gemini Decisions API alternatives
DeepSeek App API: low-cost search, file reading and voice for AI apps.
ZeroClaw: a self-hosted agent runtime to put decisions like these to work.
Frequently asked questions
Is there a free tier?
Yes, Google AI Studio has a no-cost tier for prototyping.
Can it generate text?
No. It only picks or scores answers from options you provide.
As featured in World of AI
Google's new API scores answers instead of writing them (Nov 11, 2026)
Facts checked against Google's announcement on November 13, 2026.
Want four tools like this every weeknight? Subscribe to World of AI for free. Browse more in our AI Tools directory.