Google enters decision model market with Gemini Decisions API

Google has introduced Gemini Decisions API, an AI service designed to provide cost-effective answers to user questions. This new AI model, Gemini Decision 1, is specifically engineered to handle three types of queries: yes-no questions, multiple choice questions, and rating prompts based on a user-defined scale. Compared to traditional language models, Gemini Decision 1 is five times less expensive. It can process up to 256,000 tokens of text, audio, images, video or a combination of these formats in each prompt.

Unlike most Large Language Models (LLMs), Gemini Decision 1 generates concise responses, typically "yes" or "no," along with a confidence score. This feature aids applications in determining whether to utilize the output. Decision models, a type of neural network, are known for their ability to swiftly respond to straightforward questions and are simpler to incorporate into applications. This is because they provide structured, brief answers, simplifying the process of error-checking and filtering superfluous data.

Google asserts that Gemini Decision 1's performance is nearly five times more efficient than its entry-level counterpart, Gemini 3.5 Flash-Lite LLM. Tests showed that the model outperformed Google's competitors, Jev and OpenAI's Decisions API, on seven out of nine benchmarks. However, Jev was slightly more cost-effective. Access to Gemini Decision 1 is available via the Gemini Decisions API at a rate of $0.10 per million input tokens, while output tokens are provided free of charge.

The API is also accessible through a limited free tier in Google AI Studio, a platform for developers to experiment with Google's models.