Google launches Gemini Decisions API for fast multimodal choices
Google has introduced the Gemini Decisions API, an endpoint powered by its Gemini Decision 1 model for making fast, closed-set choices across text, images, audio, video and PDFs. The service can evaluate up to 1,000 questions against one shared input and return probabilities and confidence scores.
- Gemini Decision 1 evaluates up to 1,000 questions in one request.
- The API handles text, images, audio, video and PDFs.
- Google reports 150-250 ms end-to-end latency.
- Pricing is $0.10 per 1M input tokens, with $0 for output tokens.
- An API no-cost tier is available in Google AI Studio for prototyping.
One multimodal input, up to 1,000 questions
The Gemini Decisions API is designed for bounded decisions such as tool selection, policy checks, model routing and context grading. Developers can submit a shared multimodal input of up to 256K tokens, then ask as many as 1,000 independent questions in a single request.
The questions can combine three decision types: yes/no checks, multiple choice and ordered rubric scores. Multiple-choice questions support between 2 and 255 options, while rubric questions return a continuous weighted score and probabilities for each level.
Each answer includes a confidence score from 0 to 1. Google says this allows applications to send ambiguous cases to a fallback system or a human reviewer. Inputs use the same Content block format as the Interactions API, and developers can combine modalities, such as a call recording with account metadata or a scanned claim PDF with damage photos.
Designed for low latency and lower decision costs
Unlike a generative model that produces answer text token by token, Gemini Decision 1 scores candidate options in a single forward pass. Google reports 150-250 ms end-to-end latency, including 50-100 ms of median model compute, across multimodal inputs and contexts up to 16K tokens. The company says latency remains nearly flat from 1 to 100 questions per request.
Gemini Decision 1 is priced at $0.10 per 1M input tokens, with zero output tokens billed. Google says the model and Decisions API were nearly 5x cheaper across its shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API. The company also says shared input is billed once per request, while each question adds only its own instruction and criteria tokens.
Google says Gemini Decision 1 leads across benchmarks covering agent routing, reasoning, ranking and multimodal understanding, with the lowest calibration error. These are company-reported benchmark claims.
Google teams are already using the API
Google says Gemini Notebook uses the Decisions API to route queries to the appropriate model tier and thinking level. According to the company, this can serve everyday queries with up to 40% lower latency while reserving deeper reasoning models for complex tasks.
Stitch uses the API in a rapid canvas experiment to generate playable product-flow wireframes and apply voice or tap edits. CC from Google Labs uses it to select emoji reactions from a palette of more than 200 options. Google Search uses the Decision API as a safety layer for MCP tools, detecting adversarial behavior and content against adversarial evaluations.
Developers can begin prototyping in Google AI Studio through an API no-cost tier. Google has also published Python, JavaScript and REST guides, with integrations rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode. Support is coming soon to Gemini in Google Cloud.
