Google Launches Gemini Decisions API for Multimodal AI
Google released the Gemini Decisions API to provide businesses and developers with a high-speed, cost-efficient tool for multimodal AI decision-making.
Google launched the Gemini Decisions API, powered by the Gemini Decision 1 model, to enable businesses and developers to perform multimodal AI decision-making with increased speed and lower costs. The system differs from standard generative AI by evaluating up to 1,000 independent questions—such as rubric-based scoring, multiple-choice selections, and yes-or-no assessments—against a shared input of audio, video, text, images, or PDFs in a single processing step.
The API delivers end-to-end response times between 150 and 250 milliseconds. Google priced the service at $0.10 per million input tokens and eliminated charges for output tokens.
Google has already integrated the technology into several internal products, using it as a safety layer for Google Search, for query routing in Gemini Notebook, and for product design in Stitch. This release follows the company's recent expansion of the Gemini 4 Carbon model, which focuses on cybersecurity tasks and complex reasoning.
