Google has unveiled the Gemini Decisions API, an advanced artificial intelligence tool crafted to streamline decision-making for businesses and developers. This API harnesses the power of the Gemini Decision 1 model, enabling the evaluation of up to 1,000 questions in a single request that can involve various formats such as text, images, audio, video, and PDFs.
Distinct from traditional AI models that process responses incrementally, the Gemini Decisions API assesses multiple options simultaneously, providing probability estimates and confidence scores to guide decision-making. This capability can be applied to a variety of tasks, including tool selection, policy verification, query routing, and data analysis.
Google claims that the API offers swift response times, ranging from 150 to 250 milliseconds, and is priced from $0.10 per million input tokens. Notably, there is no charge for output tokens, and shared input is billed just once, irrespective of the number of questions evaluated.
The API’s design supports different decision types—yes-or-no, multiple-choice, and ranking based on specific criteria. Its multimodal processing can handle up to 256,000 tokens of combined input, which makes it highly suitable for complex business scenarios.
Already integrated into products like Gemini Notebook, Stitch, and Google Search, the technology aids in optimizing query routing, enhancing product design, and ensuring content safety. This launch marks a significant expansion of Google’s AI ecosystem, offering a more efficient and cost-effective solution for automated decision-making tasks to developers and businesses alike.