Google launches Gemini Decisions API for calibrated multimodal scoring
Google has introduced the Gemini Decisions API, powered by Gemini Decision 1, a model designed for closed-set decisions rather than text generation. It evaluates up to 1,000 questions against one shared input and returns calibrated probabilities. Supported inputs include text, images, audio, video and PDFs, while question formats cover yes/no decisions, selections among 2 to 255 options, and ordered rubric scores.
The API is available through the Google Developer API v1beta and is hosted by Google; there are no open weights or self-hosting options. Its context limit is 262,144 tokens. Google reports typical end-to-end latency of 150–250 milliseconds for inputs up to 16,000 tokens, with question scoring performed in parallel. Input costs $0.10 per 1M tokens, output tokens are free, and the shared input is billed once per request.
In Google’s reported tests, Gemini Decision 1 led seven of nine accuracy benchmarks and recorded a top-1 calibration error of 0.0250. It scored 92.2% on the public JevBench split and 58.7% Macro-F1 on Amazon ESCI. The service currently lacks prompt caching and Batch API support. A free tier is available in Google AI Studio, and integrations with several developer platforms are being introduced.
Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.