Google Launches Gemini Decisions API for Fast Model Routing and Policy Checks

Google announced on its blog the Gemini Decisions API, a new endpoint powered by the Gemini Decision 1 model. It evaluates up to 1,000 yes or no, multiple choice and rubric scoring questions in one request without generating text token by token.

The API accepts shared inputs across text, images, audio, video and PDFs, with context of up to 256,000 tokens. It returns probabilities and confidence scores for supplied options, supporting tasks such as model routing, tool selection, policy checks and context grading.

Google reports total latency of 150 to 250 milliseconds, including median model processing time of 50 to 100 milliseconds for contexts up to 16,000 tokens. Gemini Decision 1 costs $0.10 per million input tokens, with no output token charge, and is available through the Gemini API in Google AI Studio.

Google says teams inside the company already use it. Gemini Notebook routes incoming queries to a model tier with it, serving everyday queries with up to 40% lower latency, and Google Search uses it as a safety check for MCP tools. On Google's own benchmark suite, the company says Decision 1 is nearly five times cheaper than Gemini 3.5 Flash-Lite. RuntimeWire and AlphaSignal describe it as a model for the routing and policy layer that sits in front of larger generative models.