Google has launched the Gemini Decisions API, which performs decision scoring in a single forward pass using the Decision1 model, with end-to-end latency of 150 to 250 milliseconds. It supports multimodal inputs and diverse question formats, with普惠 pricing.Author and source: AIBase
The evolution of artificial intelligence technology is accelerating from mere "content generation" toward "efficient decision-making." Google has recently launched the new Gemini Decisions API, a dedicated endpoint powered by its underlying Gemini Decision1 model, designed to enable AI agents to make rapid, predefined decisions during routine operations—eliminating the need for the cumbersome process of generating lengthy text responses in traditional models.
Core Features and Operation Mechanism: Efficient Scoring via Single Forward Pass
At the core technical architecture level, Gemini Decision1 operates fundamentally differently from traditional large models. Instead of generating responses token by token, it evaluates predefined options in a single forward pass, directly returning corresponding probabilities and confidence scores.
Thanks to this efficient design, the end-to-end latency of the interface is significantly reduced to between 150 and 250 milliseconds, meeting the real-time response requirements of agents in high-frequency interaction scenarios.
In terms of specific input and invocation formats, the Gemini Decisions API demonstrates exceptional multimodal processing capabilities. Developers can evaluate up to 1,000 independent questions in a single API call using shared multimodal inputs that include text, images, audio, video, and PDFs. Supported invocation formats include:
Quick "yes or no" check.
Multiple-choice question with up to 255 options.
Score request based on predefined evaluation criteria.
Inclusive Pricing and Application Prospects
In terms of pricing strategy, Google has set a highly competitive standard for this API: only $0.10 per million input tokens.
Industry analysts note that as large models become deeply integrated across industries, agents executing complex workflows in the background often require numerous lightweight decisions and route selections. The launch of the Gemini Decisions API, by bypassing the intermediate step of text generation, significantly reduces computational costs and response latency, providing a robust foundational infrastructure for building more seamless and responsive complex AI agent applications.

