Overview
- Google introduced the Gemini Decisions API on Wednesday, November 11, 2026, and made the gemini-decision-1 model available in Google AI Studio with a free prototyping tier.
- The API accepts a single multimodal input—text, images, audio, video, or PDFs—up to 256,000 tokens and evaluates up to 1,000 independent yes/no, multiple-choice, or rubric-scoring questions in one forward pass.
- Google reports 150–250 millisecond end-to-end latency and bills $0.10 per 1 million input tokens with no output-token charges, and it says the single-pass scorer returns probabilities and a 0–1 confidence for each answer.
- Google says teams already use the Decisions API for model routing in Gemini Notebook, rapid wireframing in Stitch, emoji selection in CC, and a Search safety layer, and ecosystem integrations with LangChain, OpenRouter, Vercel and others are rolling out.
- Buyers should test latency, calibration, and billing on their own workloads because Google’s performance and cost comparisons are self-reported and the space now includes similar decision endpoints from OpenAI and Microsoft.