NEW DELHI, Nov 12: Google has introduced a new artificial intelligence tool designed to make automated decision-making faster, more efficient and less expensive, expanding its efforts to strengthen its position in the rapidly evolving AI industry.
The technology giant announced the Gemini Decisions API, powered by its newly developed Gemini Decision 1 model, which allows developers to evaluate up to 1,000 questions within a single request. The system is designed to handle specific decision-making tasks rather than generate lengthy conversational responses, potentially making it useful for businesses, software developers and organisations building AI-powered applications.
The announcement represents another step in the development of specialised artificial intelligence systems that can perform targeted operations with greater speed and efficiency. As companies increasingly integrate AI into their products and services, the ability to process information and make routine decisions quickly has become an important requirement.
Unlike conventional generative AI models, which generally produce responses by generating tokens sequentially, Gemini Decision 1 evaluates predefined options and returns probabilities and confidence scores. This approach is intended to reduce unnecessary processing while allowing applications to receive structured results.
Google said the new API supports three principal decision-making functions: yes-or-no assessments, multiple-choice selections and scoring based on predefined criteria. Developers can use these capabilities for tasks such as checking whether certain conditions have been met, selecting an appropriate action or evaluating information against specific standards.
One of the most significant features is its ability to process different forms of information within a single request. The system supports text, images, audio recordings, video and PDF documents, allowing developers to combine multiple sources when evaluating a particular situation.
For example, a business could analyse a customer-service recording alongside account information to determine the appropriate response. Similarly, an insurance company could examine a scanned claim document together with photographs of damaged property to assess whether the submitted information meets specified requirements.
The technology can also support content moderation, automated quality checks, document classification and the selection of suitable AI tools for different operations.
According to Google, the model can handle inputs containing up to 256,000 tokens and evaluate as many as 1,000 independent questions in one request. The company has also highlighted response speeds of approximately 150 to 250 milliseconds, making the system suitable for applications where delays could affect the user experience.
Another important aspect of the launch is its pricing structure. Google has set the cost at USD 0.10 per million input tokens, with no charges for output tokens. The company said the shared input is billed once per request, regardless of how many questions are evaluated against it.
This pricing arrangement could make the service attractive to developers managing applications that repeatedly examine the same information through multiple checks.
Google has claimed that Gemini Decision 1 is nearly five times cheaper than Gemini 3.5 Flash-Lite with the Interactions API across its shared benchmark suite. These comparisons are based on the company’s own testing and may vary depending on the workload and implementation.
The system also provides confidence scores alongside its results. These scores can help applications identify uncertain decisions and transfer complicated cases to human reviewers or alternative AI models.
Such a feature could prove particularly useful in sectors where automated decisions require additional scrutiny. Businesses dealing with financial documents, customer information, regulatory requirements and sensitive records may benefit from systems that can distinguish between straightforward cases and situations requiring human attention.
Google has also identified AI model routing as an important application. As organisations increasingly use multiple AI models with different capabilities and costs, selecting the appropriate system for each incoming request can become a complicated task.
The Decisions API could help automate this process by assessing the nature of a query and identifying which model or tool should handle it. This could reduce unnecessary use of expensive computing resources while improving response times.
The company said its technology is already being used in several internal and experimental applications. Gemini Notebook uses the system to route incoming queries towards suitable model tiers, while Google’s Stitch design experiment uses it to support rapid design interactions.
The technology is also being incorporated into Google’s Search safety processes to identify adversarial behaviour and potentially harmful content.
The launch comes amid growing competition among major technology companies to develop AI systems that are not only more powerful but also cheaper and more practical for everyday business applications.
While large language models have attracted considerable attention for their ability to generate text, images, computer code and other forms of content, companies are increasingly exploring specialised systems that can perform individual tasks without requiring extensive generation.
For developers, the distinction could be important. Many software applications do not need an elaborate written response every time they process information. They may simply require an assessment, a classification, a score or a selection from a predefined list.
By focusing on these requirements, Google is attempting to address a segment of the AI market where speed, predictable output and operating costs are particularly important.
The company has made Gemini Decision 1 available through the Gemini API in Google AI Studio, where developers can access the service and begin experimenting with its capabilities. Google has also announced integrations involving development platforms and frameworks, including OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode.
The broader availability of such integrations could help developers incorporate the technology into existing software projects without having to build an entirely new infrastructure.
However, the introduction of specialised decision-making systems also raises questions about reliability, transparency and accountability. Confidence scores can help identify uncertain outcomes, but they do not automatically guarantee that every decision is correct.
Organisations deploying these tools will still need to establish appropriate testing procedures, monitor performance and ensure that important decisions are reviewed when necessary.
There are also practical considerations involving the quality of information supplied to the system. Incorrect, incomplete or misleading input could affect the reliability of the results, particularly when decisions involve several sources of information.
For this reason, human supervision is likely to remain important in applications involving financial assessments, regulatory compliance and other sensitive operations.
Google’s latest announcement reflects a wider shift in the artificial intelligence industry towards specialised models designed for specific business requirements. Rather than relying exclusively on increasingly large generative systems, technology companies are exploring ways to divide complex workflows into smaller operations that can be completed more efficiently.
Whether Gemini Decision 1 achieves widespread adoption will depend on its performance in real-world applications, the reliability of its assessments and the extent to which developers find its pricing and integration options competitive.
Nevertheless, the launch adds another capability to Google’s expanding AI ecosystem and highlights the growing importance of fast, structured and cost-conscious artificial intelligence solutions.