The evolution of artificial intelligence technology is accelerating from simply “generating content” toward “efficient decision-making.” Google has recently officially launched the new Gemini Decisions API, a new endpoint powered by its underlying Gemini Decision1 model. It is designed to help AI agents make lightning-fast predefined decisions during day-to-day operations, completely eliminating the cumbersome process of generating lengthy text responses under the traditional approach.

Core Features and Operating Mechanism: Efficient Scoring in a Single Forward Pass

At the underlying technical architecture level, Gemini Decision1 is fundamentally different from how traditional large models operate. The model does not generate responses token by token. Instead, it directly scores predefined options in a single forward pass and ultimately returns the corresponding probability and confidence scores.

Thanks to this efficient design, the endpoint’s end-to-end latency is greatly reduced to between 150 and 250 milliseconds, meeting the real-time response needs of agents in high-frequency interaction scenarios.

In terms of specific input and invocation formats, the Gemini Decisions API demonstrates strong multimodal processing capabilities. Developers can evaluate up to 1,000 separate questions in a single API call against shared multimodal input that includes text, images, audio, video, and PDFs. The specific supported invocation formats include:

Quick “yes or no” checks.

Multiple-choice questions with up to 255 options.

Scoring requests based on predefined scoring criteria.

Affordable Pricing and Application Prospects

In terms of pricing strategy, Google has set a highly competitive rate for the API: only $0.10 per million input tokens.

Industry analysts point out that as large models become deeply embedded across industries, agents often need to make numerous lightweight judgments and routing choices when executing complex workflows in the background. By skipping the intermediate step of text generation, the launch of the Gemini Decisions API not only greatly reduces computational costs and response latency, but also provides a solid underlying infrastructure for building smoother, more responsive complex AI agent applications.