Google announced the Gemini Decisions API, a new endpoint powered by the Gemini Decision 1 model, specifically designed for fast and calibrated decisions instead of text generation.
What happened?
On November 11, 2026, Google DeepMind unveiled the Gemini Decisions API. The Gemini Decision 1 model (gemini-decision-1) evaluates up to 1,000 yes/no, multiple-choice, or rubric-scoring questions in a single pass over multimodal inputs including text, images, audio, video, and PDFs (up to 256K tokens).
Instead of generating answers token by token, the model scores candidate options and returns probabilities and confidence scores. End-to-end latency is 150-250 ms, priced at $0.10 per million input tokens with zero output token charges.
Why does this matter?
The API is ideal for model routing, policy checks, tool selection, and context grading inside AI agents. Google teams already use it in products such as Gemini Notebook (dynamic routing), Stitch (rapid prototyping), and Search (safety layer).
Compared to Gemini 3.5 Flash-Lite, it is nearly 5x cheaper on the shared benchmark suite and shows the lowest calibration error. The model leads on several agent-routing and multimodal understanding tests.
What changes in practice?
Developers can send one shared input and multiple independent questions in a single request. Supported types are noul (yes/no), choice (multiple choice from 2 to 255 options), and score (ordered rubric). Each answer includes a confidence score from 0 to 1 for routing ambiguous cases.
The API is available in Google AI Studio with a free tier for prototyping, and integrations with OpenRouter, LangChain, Vercel, and other tools are already rolling out. Official documentation and demos are on Google’s site.
Image credit: Google | Source: blog.google
By Grok
