Google Gemini Decisions API: 1,000 Yes/No Checks for 60 Cents
In short
- Google launched the Gemini Decisions API: one call, up to 1,000 yes/no, choice or scoring questions about the same text, image, audio, video or PDF.
- It costs $0.10 per 1M input tokens. Output tokens cost nothing. Answers come back in 150–250 ms.
- Sorting 10,000 customer emails costs about 60 cents instead of days of manual work.
Google just made classification almost free. On November 11 it launched the Gemini Decisions API. You don’t ask a chatbot to write an answer. You hand it a document and a list of questions, and it returns probabilities. Is this email a complaint? Is this invoice complete? Does this photo break your policy? Yes or no, with a confidence score.
What happened
The new endpoint runs on a model called Gemini Decision 1. It takes one input (text, image, audio, video or PDF, up to 256K tokens) and up to 1,000 independent questions about it. Three question types: yes/no, multiple choice (2 to 255 options) and a rubric score.
It doesn’t write text. It scores your options in a single pass and returns probabilities from 0 to 1. That’s why it’s fast: Google quotes 150–250 ms end to end, and the speed barely changes between 1 and 100 questions.
How much does it cost? $0.10 per million input tokens. Output tokens are free, and there are no thinking-token charges. The shared document is billed once per request, and each question adds only its own few tokens. There’s a free tier in Google AI Studio for testing. The endpoint is still v1beta.
How to use it
- Inbox triage. One email, five questions: urgent? complaint? refund? spam? which department? Route it automatically.
- Policy checks. Run every product listing or review through a list of rules before it goes live.
- Lead scoring. Score each inquiry against your own rubric and send only the top ones to a human.
- Model routing. Let a cheap call decide which requests deserve an expensive model.
Do the math: 10,000 emails a month, about 600 tokens each including the questions. That’s 6M tokens, so about $0.60 (roughly €0.55). Doing it by hand at one minute per email is 167 hours. At €10 an hour that’s €1,670. That’s my assumption, not a Google number.
Google also lists integrations rolling out for OpenRouter, LangChain, Vercel and LiteLLM. If you build in Make or n8n, a plain HTTP call is enough.
What it means
Most business AI work isn’t writing. It’s deciding. Sort this, approve that, flag the rest. Until now you paid for a full language model to answer yes or no. Google says this is nearly 5x cheaper than Gemini 3.5 Flash-Lite on its own benchmark suite. That’s a vendor claim, so test it on your own data.
The catch: it’s a beta endpoint, and a confidence score is not a guarantee. Keep a human on anything that touches money or customers.
Read also: Gemini Omni 1.2 Flash: 15-Second AI Video at $0.07 a Second · OpenAI’s Official CLI: Automate API Tasks From One Command · Edit Photos by Describing the Change for €0 in Gemini
Ivan’s Take
This is the boring kind of AI that actually saves money. I’d test it first on incoming customer messages and listing checks, where one wrong answer is cheap to catch. Run 500 real examples against your own judgment before you trust it with anything bigger.
Source: Google
HawkValue
If it doesn’t save you time or money, you won’t see it here
- Facts, not hypeEvery price and number is checked against the official source
- Always the real priceYou see what it costs before you click
- I say when it’s weakEven when it’s OpenAI or Google
