Google has unveiled its new Gemini Decisions API, an artificial intelligence tool aimed at enhancing the speed and affordability of decision-making for businesses and developers. The API is powered by the Gemini Decision 1 model and can process up to 1,000 questions that involve various data types, including text, images, audio, video, and PDF documents, all within a single request.
Unlike traditional AI models that generate responses sequentially, the Gemini Decisions API evaluates multiple options simultaneously and provides probability estimates and confidence scores. This system is designed for tasks such as selecting tools, checking policies, routing queries, and analyzing information, offering response times ranging from 150 to 250 milliseconds.
Google has set the pricing model for the API at $0.10 per million input tokens, with no charges for output tokens. Shared inputs are billed only once, regardless of the number of questions evaluated, making it a cost-effective solution for businesses. The API supports different types of decision-making processes, including yes-or-no decisions, multiple-choice selections, and ranking options based on specified criteria. Its multimodal capabilities also allow for the processing of up to 256,000 tokens of combined input, catering to complex business applications.
Already integrated into several Google products such as Gemini Notebook, Stitch, and Google Search, the technology is being used to improve query routing, product design, and content safety. The launch of the Gemini Decisions API represents an expansion of Google’s AI ecosystem, offering a more efficient and economical option for developers and businesses seeking to handle individual evaluations and automated decision-making tasks.