Google has unveiled the Gemini Decisions API, a cutting-edge artificial intelligence tool aimed at enhancing decision-making processes for businesses and developers. This new API, powered by the Gemini Decision 1 model, can simultaneously evaluate up to 1,000 questions involving various data types such as text, images, audio, video, and PDF documents in a single request.
Unlike conventional AI models that provide responses incrementally, the Gemini Decisions API processes multiple options in one step, offering probability estimates and confidence scores. This functionality can be applied to a range of tasks including tool selection, policy checks, query routing, and information analysis.
Google claims that the API delivers rapid response times of approximately 150–250 milliseconds, and it is priced economically at $0.10 per million input tokens, with no charge for output tokens. Additionally, when the same input is shared across multiple questions, it is billed only once, making it a cost-effective solution for complex evaluations.
The API supports various decision types, including yes-or-no, multiple-choice, and ranking based on specified criteria. Its multimodal capabilities allow it to handle up to 256,000 tokens of combined input, making it versatile for complex business applications.
Already integrated into several Google products such as Gemini Notebook, Stitch, and Google Search, the API aims to improve query routing, product design, and content safety. This launch marks a significant expansion of Google’s AI ecosystem, offering developers and businesses a more efficient and affordable tool for automated decision-making tasks.