AI tools ·
Google's Gemini Decision 1 answers 1,000 questions in one pass — for $0.10 per million tokens
Google launched the Gemini Decisions API on November 11, powered by a new model called Gemini Decision 1. Instead of writing text, the model scores up to 1,000 yes/no, multiple-choice, or rubric questions about a single input — text, images, audio, video, or a PDF — in one pass, returning a probability and a confidence score for each answer in 150 to 250 milliseconds. It costs $0.10 per million input tokens, and output tokens are not billed at all, because the model produces none.
The API is built for the small, bounded choices inside bigger AI systems: picking which tool an agent should call, routing a question to the right model, checking a policy, or grading content. Google says it already uses the Decisions API inside its own products — Gemini Notebook routes queries with it, cutting latency up to 40% on everyday questions, and Google Search uses it as a safety layer for its MCP tools. Developers can try it free in Google AI Studio; it runs on the Gemini API's v1beta endpoint, with integrations rolling out to OpenRouter, LangChain, Vercel, and LiteLLM. Support in Google Cloud is coming soon.
The launch puts Google in a fast-growing category. OpenAI shipped its own Decisions API in September, Microsoft followed with Microsoft-Decision-1, and startup TypeSafe has been selling the same idea with Jev since September. Google's edge is price and multimodality: one shared input, a thousand questions, a dime per million tokens — and a free tier to try it.
Why it matters
Most AI spending comes from models writing long answers token by token. A model that only scores choices — and charges only for input — turns judgment into a cheap utility: spam checks, content grading, agent routing, and safety checks can run on every request instead of a sample. For small teams, the free tier in AI Studio means the cheapest decision layer in the industry is now a signup away.
Sources
- Google — Gemini Decisions API: Fast, calibrated multimodal decisions
- MadRobot — Google's Gemini Decisions API answers without writing
- MLQ.ai — Google launches Gemini Decision 1, a model that scores up to 1,000 questions in one pass
Spot an error? We correct quickly and note it. Contact us via the contact page.