Alphabet Inc.’s Google has launched Gemini Decisions API, a specialized artificial intelligence (AI) service designed to streamline software integration and drastically cut computing costs for enterprise developers.
Powered by a new model, Gemini Decision 1, the service trades the broad conversational abilities of traditional large language models for rapid, structured outputs at a fraction of the price.
Unlike standard LLMs that generate lengthy, free-flowing text, Gemini Decision 1 is strictly limited to answering three specific types of prompts: yes-or-no questions, multiple-choice queries, and numerical ratings on a user-defined scale. For instance, a cybersecurity team could use the tool to evaluate an antivirus alert and assign a severity score from one to 10.
To process a request, developers feed the model a question alongside a relevant dataset containing up to 256,000 tokens of text, audio, images, or video. Instead of prose, the model returns a concise response—such as a simple “yes” — paired with a confidence score. This metric allows downstream applications to programmatically determine whether an automated answer is reliable enough to use without human intervention.
According to Google, this narrow focus yields major performance and financial benefits. The company states that Gemini Decision 1 is nearly five times cheaper than its entry-level Gemini 3.5 Flash-Lite model. By scoring options directly rather than generating text token by token, the model maintains a median compute latency of 50 to 100 milliseconds—or 150 to 250 milliseconds end-to-end—even when handling up to 100 questions in a single request. Because the outputs are brief and predictable, developers spend less time filtering redundant text or checking for formatting errors.
The release places Google directly into a rapidly heating market for specialized “decision models.” Startup TypeSafe AI Inc. pioneered the category in mid-September with its model Jev, rapidly attracting a million users and securing $870 million in funding at a $7.5 billion valuation weeks later. OpenAI Group PBC followed with its own competing Decisions API last month.
In internal testing, Google reported that Gemini Decision 1 outperformed both Jev and OpenAI’s offering across seven of nine benchmarks, though Jev retained a slight edge in overall cost efficiency.
Gemini Decision 1 is available immediately via the Gemini Decisions API at 10 cents per million input tokens, with no charges for output tokens. Developers can also test the model through a limited free tier on Google AI Studio.
