Google launches new Gemini Decisions API for better decision-making in AI applications

Google has introduced the Gemini Decisions API, powered by Gemini Decision 1, its first dedicated decision model. Instead of generating text responses, the model assigns probability scores to predefined options. Developers can evaluate up to 1,000 questions in a single request, with support for binary decisions, multiple-choice selection, and rubric-based scoring. The API also returns confidence scores to help determine when decisions can be automated or require further review.

Gemini Decision 1 supports text, images, audio, and video, with a context window of up to 256,000 tokens. Google reports latency of 150 to 250 milliseconds for contexts up to 16,000 tokens, with relatively stable performance across one to 100 questions. Pricing starts at $0.10 per million input tokens, with no output-token fees. Shared input is billed only once per request, and additional questions incur charges only for their instruction and criteria tokens.

Google already uses the technology in Gemini Notebook to route queries based on complexity, reportedly reducing latency for everyday queries by up to 40%. Google Search also uses it to detect adversarial behavior involving Model Context Protocol (MCP) tools. The API is available through Google AI Studio, with developer framework integrations rolling out and Google Cloud support planned for a later release.