Gemini Decisions API (blog.google)

Google has unveiled the Gemini Decisions API, an innovative new endpoint for making rapid, calibrated multimodal decisions using the Gemini Decision 1 model. This API enables users to analyze text, images, audio, video, and PDFs in a single request, evaluating up to 1,000 independent questions while processing inputs of up to 256K tokens. Unlike conventional models that generate outputs token by token, Gemini Decision 1 achieves impressive speeds with an end-to-end latency of just 150-250 ms, allowing for efficient decision-making with minimal costs of $0.10 per million input tokens and no fees for output tokens. This release is significant for the AI/ML community as it enhances capabilities for dynamic model routing, rapid design iterations, and effective safety layers in applications. Organizations can employ the Decisions API for diverse tasks, such as automating responses in Google Search or driving user interface design in Stitch, with improved speed and reduced latency. By facilitating seamless multimodal processing, Gemini Decision 1 not only lowers operational costs—up to 5x cheaper than previous iterations—but also empowers developers to create advanced applications without extensive resource allocation, marking a notable advance in decision-making technologies in AI.