Google has launched the Gemini Decisions API, capable of evaluating up to 1,000 multimodal queries at once with latency at the 150-millisecond level, with free outputs now available on AI Studio.Author and source: AIBase
Let AI take over quick yes-or-no and multiple-choice decisions—Google has reduced latency to milliseconds. On November 11, Google announced the launch of the Gemini Decisions API, powered by the new Gemini Decision 1 model, specifically designed for scenarios requiring repeated yes/no questions, selection tasks, and scoring.
Feed a thousand questions at once; forward computation assigns scores directly.
Developers upload multimodal materials composed of text, images, audio, video, or PDFs (up to 256K tokens) and can ask up to 1,000 unrelated questions. Gemini Decision 1 performs just one forward pass to assign probability and confidence scores to each option—eliminating the need for hundreds of separate calls as in traditional workflows. This capability is ideal for tool selection, strategy verification, model routing, and context quality assessment: for example, feeding in a call recording alongside account details, or submitting a scanned claim form with photos of damage, allowing the model to verify everything in a single pass.
Fast and affordable, with adjustable resolution
In terms of speed, end-to-end processing latency ranges from 150 to 250 milliseconds. Pricing is also competitive: input costs just $0.10 per million tokens (approximately ¥0.67 RMB), while output tokens are completely free. Developers can also manually adjust the resolution of images and videos to balance visual detail against token consumption. The API is now live on Google AI Studio, offering a free tier for prototype development.