Google teaches AI to make decisions without excess text — Gemini Decisions API will save on tokens
Google has introduced the Gemini Decisions API and the Gemini Decision 1 model for fast verifications and option selection. The service processes multimodal data and up to 1,000 questions per request.
The new interface is available via the Gemini API in Google AI Studio. Gemini Decision 1 does not generate responses sequentially but evaluates the given options in a single pass and returns probabilities and a confidence score. The total processing latency is 150–250 ms, of which 50–100 ms is spent on model computations.
The model works with text, images, audio, video, and PDFs. A single request can include up to 256,000 tokens and up to 1,000 independent questions. The API supports yes/no checks, selecting one option from a list of 2 to 255 items, and evaluation on an ordered scale. For each result, a confidence score between 0 and 1 is returned, allowing ambiguous cases to be flagged for additional human review.
Google stated that Gemini Decision 1 demonstrated the best results in a set of tests for routing, reasoning, ranking, and multimodal understanding, as well as the lowest calibration error. According to the company, in these tests, the combination with the Decisions API is nearly five times cheaper than Gemini 3.5 Flash-Lite with the Interactions API.
Gemini Notebook uses the API to select the model and reasoning level: Google claims this reduces latency for standard requests by up to 40%, reserving more complex models for difficult tasks. Stitch employs it for rapid prototyping, while Google Search uses it to detect malicious behavior when working with MCP tools.
The price is $0.10 per 1 million input tokens, with output tokens not billed. The total input data is charged once per request regardless of the number of questions, and a free tier is available for prototyping in Google AI Studio.
