models

Google's Gemini Decisions API: Fast, Calibrated Multimodal Decisions

Summarized by AI from reporting by Google AI Blog, published under our editorial policy.

Google has launched the Gemini Decisions API, a new tool that can evaluate up to 1,000 yes/no, choice, and rubric-scoring questions in a single pass across text, images, audio, video, and PDFs. It is designed for fast, probabilistic choices with low latency and cost.

A screenshot of the Gemini Decisions API interface, showcasing its multimodal decision-making capabilities.

Key takeaways

  • The Gemini Decisions API can evaluate up to 1,000 questions in a single pass across text, images, audio, video, and PDFs.
  • Gemini Decision 1 processes multimodal inputs in a single forward pass with 150-250 ms end-to-end latency.
  • The API is nearly 5x cheaper across the shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API.
  • Gemini Decision 1 is priced at $0.10 per 1M input tokens with zero output tokens billed.
  • The API supports yes/no checks, multiple-choice selections, and rubric scoring with confidence scores.

Google has introduced the Gemini Decisions API, powered by the new Gemini Decision 1 model. This tool allows users to evaluate up to 1,000 yes/no, choice, and rubric-scoring questions in a fast, single pass across various data types, including text, images, audio, video, and PDFs. The API is designed to provide quick, probabilistic decisions with minimal latency and cost. This is a significant step forward in multimodal decision-making, building on Google's existing AI capabilities and offering a more efficient and cost-effective solution for a wide range of applications.

What the Gemini Decisions API Actually Does

The Gemini Decisions API is built to handle bounded choices like tool selection, policy checks, model routing, and context grading. Users can pass one multimodal input (up to 256K tokens) along with up to 1,000 independent questions. Instead of generating answer text token by token, the Gemini Decision 1 model scores candidate options in a single forward pass, returning probabilities and confidence scores in 150-250 ms end-to-end latency. The cost is $0.10 per 1M input tokens with zero output tokens billed. This approach is significantly faster and more cost-effective than traditional methods, which often require multiple passes and higher computational costs.

The API supports three decision types: yes/no checks, multiple-choice selections, and rubric scoring. For yes/no checks, it estimates the probability that a condition holds, with optional true and false description criteria. For multiple-choice questions, it selects from 2 to 255 choices, returning the winner and normalized probabilities for every candidate. For rubric scoring, it grades inputs across an ordered rubric, returning a continuous weighted score and per-level probabilities. This flexibility makes the API suitable for a wide range of applications, from content grading to policy enforcement.

Key Features and Performance

Gemini Decision 1 leads in benchmarks spanning agent routing, reasoning, ranking, and multimodal understanding, with the lowest calibration error. It processes text, images, audio, video, and PDFs directly in a single forward pass. Users can combine modalities in one shared input, such as pairing a call recording with account metadata or a scanned PDF claim with damage photos. The API supports three decision types: yes/no checks, multiple-choice selections, and rubric scoring, each with a confidence score.

The API's low latency and cost make it a practical tool for real-time decision-making. With a median latency of 50-100 ms of model compute (150-250 ms end-to-end), it remains nearly flat from 1 to 100 questions per request. This efficiency is achieved through a single forward pass, eliminating the need for token-by-token decoding. The API is nearly 5x cheaper across the shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API, making it an attractive option for developers and businesses looking to integrate advanced decision-making capabilities into their applications.

Why It Matters for Everyday Users

The Gemini Decisions API can significantly impact various applications, from dynamic model routing in Gemini Notebook to rapid design iteration in Stitch. It can also be used for emoji selection in CC from Google Labs and as a safety layer for MCP tools in Search. The API's low latency and cost make it a practical tool for real-time decision-making in everyday applications.

For example, Gemini Notebook uses the Decisions API as a real-time router to instantly direct incoming queries to the optimal model tier and thinking level for each request. This classification of intent in milliseconds serves everyday queries with up to 40% lower latency while reserving deep reasoning models for complex tasks. Similarly, the new rapid canvas experiment in Stitch uses the Decisions API to generate playable wireframes of a product flow in seconds and to apply edits made by voice or tap almost instantly. This rapid iteration capability is invaluable for designers and developers looking to streamline their workflows.

The API's ability to handle multimodal inputs makes it particularly useful for applications that require processing and analyzing different types of data. For instance, CC from Google Labs uses the Decisions API to pick the right emoji reaction from a palette of more than 200 options in a fraction of a second, ensuring that reactions appear instantly without adding latency. This real-time decision-making capability enhances user experience and engagement.

What You Can Do Today

You can start building with the Gemini Decisions API today by accessing your API key in Google AI Studio. Review the documentation for Python, JavaScript, and REST guides, and try the demo apps like Doorstep Parcels, Speed Sketch, and Find the Zoo Keeper. These demo apps provide practical examples of how the API can be used to solve real-world problems, from routing parcels to identifying objects in images. The API's user-friendly design and comprehensive documentation make it accessible for developers of all skill levels.

Additionally, the API offers a no-cost tier for prototyping in Google AI Studio, allowing you to experiment and test its capabilities without incurring costs. This makes it an ideal tool for startups and small businesses looking to integrate advanced decision-making capabilities into their applications without a significant upfront investment. As you become more familiar with the API, you can scale up your usage and take advantage of its full potential.

Frequently asked

Is the Gemini Decisions API free to use?
The Gemini Decisions API offers a no-cost tier for prototyping in Google AI Studio, but it is priced at $0.10 per 1M input tokens for full use.
What types of questions can the Gemini Decisions API evaluate?
The API can evaluate yes/no, multiple-choice, and rubric-scoring questions across various data types.
How can I start using the Gemini Decisions API?
You can start by accessing your API key in Google AI Studio and reviewing the documentation for Python, JavaScript, and REST guides.
What is the latency of the Gemini Decisions API?
The API has an end-to-end latency of 150-250 ms, with model compute latency of 50-100 ms.