Google is launching the Gemini Decisions API, a new endpoint that lets developers have up to 1,000 questions answered at once. The new Gemini Decision 1 model returns probabilities and confidence scores in 150 to 250 ms.
The Gemini Decisions API is available through the Gemini API in Google AI Studio. The endpoint is designed for decisions with a fixed set of answers, such as choosing a tool, checking policy rules, routing models and assessing context.
Developers pass in a single input, which can combine text, images, audio, video and PDFs up to 256K tokens. Up to 1,000 independent questions can be added to it.
Scoring instead of generating
Gemini Decision 1 (gemini-decision-1) does not generate answer text token by token. The model scores the options in a single pass and returns probabilities and confidence scores. According to Google, the model’s median compute time is 50 to 100 ms, and latency stays virtually the same at 1 to 100 questions per request.
There are three question types. With a yes/no question, the model estimates the probability that a condition holds. With a choice question, the model selects from 2 to 255 options and you get the normalised probabilities of each candidate. A score question assesses the input against an ordered rubric and produces a weighted score.
Each answer includes a confidence score between 0 and 1. This lets your code forward unclear cases to a fallback option or a human reviewer. Examples include a call recording combined with account data, or a scanned PDF with photos of damage.
Pricing and use within Google
Gemini Decision 1 costs $0.10 per million input tokens. Output tokens are not charged. You pay for the shared input once per request, regardless of the number of questions, and each question only adds its own instruction and criteria tokens. There are also no charges for thinking tokens, audio transcription or OCR. According to Google, the combination with the Decisions API is almost five times cheaper than Gemini 3.5 Flash-Lite with the Interactions API, measured across the same benchmark suite.
Google says the model achieves the best results on benchmarks for agent routing, reasoning, ranking and multimodal understanding, with the lowest calibration error. Internally, the company already uses the API for several products:
Gemini Notebook uses it to route incoming questions to the appropriate model and thinking level, with up to 40 percent lower latency for everyday questions. The Stitch feature rapid canvas uses it to generate playable wireframes in a few seconds. CC from Google Labs uses it to pick an emoji reaction from more than 200 options, and Google Search uses the API as a safety layer for MCP tools to detect malicious behaviour.
Availability
Developers can get started today with an API key in Google AI Studio, where a free tier for prototyping is also available. There are guides for Python, JavaScript and REST, and three demo apps: Doorstep Parcels, Speed Sketch and Find the Zoo Keeper.
Integrations are coming for OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode, alongside the google-genai SDKs. Support in Gemini in Google Cloud will follow later. Google does not give a date for this, and the company has not announced separate pricing for Belgium either.