Google announced the Gemini Decisions API on November 11, 2026. It runs on the Gemini Decision 1 model, named gemini-decision-1, and is reachable through the Gemini API in Google AI Studio. A single request can assess up to 1,000 questions of three kinds: yes/no checks, multiple choice, and rubric scoring. Those questions are answered against one shared input that may mix text, images, audio, video and PDFs.
The service returns probabilities and confidence scores for every answer. Each confidence score runs from 0 to 1, so an application can send unclear cases to a fallback or a human reviewer. Google reports 150-250 ms end-to-end latency. Input is billed at $0.10 per 1M input tokens, with no output tokens billed. Inputs can reach 256K tokens.
Google teams already use the Decisions API. Gemini Notebook routes incoming requests with it, serving everyday requests with up to 40% lower latency. CC from Google Labs uses it to pick the right emoji reaction from more than 200 options. A no-cost tier in Google AI Studio is available for prototyping. Integrations are rolling out across OpenRouter, LangChain, Vercel and LiteLLM, while Gemini in Google Cloud support is coming soon.
Key facts
- The Gemini Decisions API runs on the Gemini Decision 1 model, identified as gemini-decision-1.
- Up to 1,000 questions can be scored in a single pass against one multimodal input.
- Each answer carries a confidence score from 0 to 1.
- Input is charged at $0.10 per 1M input tokens, and no output tokens are billed.
- Gemini Notebook uses the API to route queries, with up to 40% lower latency for everyday queries.
Why it matters
Developers can start by getting an API key in Google AI Studio and following the Python, JavaScript, and REST guides. The API is meant for bounded choices such as tool selection, policy checks, model routing and context grading. Instead of producing answer text one token at a time, it scores candidate options in one forward pass. Pricing is $0.10 per 1M input tokens with zero output tokens billed.




