Google Gemini Decisions API Debuts With Support for 1,000 Questions Per Request

Google has introduced the Gemini Decisions API, which allows developers to use its new Gemini Decision 1 model to evaluate information and make quick decisions. It can assess up to 1,000 questions in a single request, using text, images, audio, video and PDF files as input. Developers can ask it to make yes-or-no decisions, choose from a list of options or score content against set criteria. The Mountain View-based tech giant sees potential uses in content moderation, AI model routing and deciding which tools an agent should use.

Google Gemini Decisions API Features

Google has launched the Gemini Decisions API for developers who need AI to make quick decisions rather than generate lengthy responses. The API runs on its new Gemini Decision 1 model and can help developers check content against specific policies, route queries to different AI models or pick the right tools for an AI agent.

Developers can submit text, images, audio, video and PDF files for the model to analyse. It can answer up to 1,000 questions about the same input in a single request. These can be yes-or-no questions, multiple-choice queries or requests to score content against a set of criteria. The model also returns a confidence score with each answer.

Gemini Decision 1 handles these tasks differently from Google's generative AI models. Rather than building a response one token at a time, it weighs the available options and returns its decisions in a single pass. Google claims this approach delivers results in 150 to 250 milliseconds.

Google has started using the API across several of its products. Gemini Notebook, for instance, uses it to pick the right model and thinking level for incoming queries. Google says this has cut latency for everyday requests by up to 40 percent. The company also uses the API in Stitch to support product design and in its CC app to pick emoji reactions. In Google Search, it helps flag attempts to misuse the service's tools.

Google Gemini Decisions API Availability, Pricing

Developers can try the Gemini Decisions API for free through Google AI Studio. Paid usage costs $0.10 (roughly Rs. 9.39) per million input tokens, while Google does not charge for output tokens. It also bills shared input only once, even when developers include multiple questions in a request. Support for OpenRouter, LangChain, Vercel and LiteLLM is rolling out.