ToolGemini

Google launches Gemini Decisions API for fast decisions

Original title: Gemini Decisions API: Fast, calibrated multimodal decisions

Plain language

What it is: Google introduced the Gemini Decisions API, powered by the new Gemini Decision 1 model. It is designed to make fast choices, like yes/no or multiple-choice questions, based on text, images, audio, video, and PDFs.
How it works: Instead of writing out answers word by word, the model scores all options at the same time in a single pass. This makes the process very fast and cheap, providing confidence scores for up to 1,000 questions at once.
Where it is available: It is available through the Gemini API in Google AI Studio.
Example: A business can use it to check a photo of a breakroom counter and a text note to instantly see if there is a spill and rate its severity.

What it is: Google launched the Gemini Decisions API and the Gemini Decision 1 model for fast, closed-set multimodal decisions.
How it works: The model processes up to 256K tokens of multimodal input and up to 1,000 independent questions in a single forward pass without token-by-token decoding. It returns probabilities and confidence scores for three decision types: yes/no (noul), multiple choice (choice), and ordered rubric (score).
Where it is available: It is accessible via the Gemini API in Google AI Studio, with integrations rolling out to platforms like LangChain, Vercel, and LiteLLM.
Example: Developers can call client.decisions.create with an image and text to evaluate questions like has_spill using the noul type and receive a confidence score.

The headline and explanation on this page are based on Google's post. The full details are in the original.

Gemini Decisions API: Fast, calibrated multimodal decisions