Google launched the Gemini Decisions API on Wednesday, a new endpoint in the Gemini API and Google AI Studio built to answer closed-set questions instead of generating text. It runs on a new model, Gemini Decision 1, and Google is charging $0.10 per million input tokens with nothing billed for output.
The design trades open-ended answers for volume and speed. A single request can carry up to 1,000 independent questions against one shared input, and the model scores the candidate options in a single forward pass rather than writing a response token by token. Every answer comes back with a confidence score between 0 and 1.
Google puts median model compute at 50 to 100 milliseconds, or 150 to 250 milliseconds end to end. The company said that latency holds across multimodal inputs and contexts up to 16K tokens and stays nearly flat whether a request carries 1 question or 100.
Key facts
- 01Google launched the Gemini Decisions API on Nov. 11, 2026, powered by a new model called Gemini Decision 1.
- 02One request can carry up to 1,000 independent yes-or-no, multiple-choice or rubric questions against a shared input.
- 03Pricing is $0.10 per 1M input tokens and $0 for output tokens; shared input is billed once per request.
- 04Google lists 150-250 ms end-to-end latency, nearly flat from 1 to 100 questions on inputs up to 16K tokens.
- 05Gemini Notebook uses the API for model routing and saw up to 40% lower latency on everyday queries, Google said.
The API supports three question formats: yes-or-no questions, choices between 2 and 255 options, and ordered scoring rubrics with 2 to 255 levels. The shared input can mix text, images, audio, video and PDFs, up to roughly 256K tokens. Google's developer documentation puts the exact context window at 262,144 tokens, shared across the input and all questions.
Billing is structured around that shared input. Google charges for it once per request regardless of how many questions are attached, and each question adds only the tokens in its own instruction and criteria. A no-cost tier is available in Google AI Studio for prototyping.
Google is positioning Gemini Decision 1 for the plumbing of agent systems: tool selection, policy checks, model routing and grading context. The company said the model leads a benchmark suite spanning agent routing, reasoning, ranking and multimodal understanding with the lowest calibration error, and that it is nearly 5x cheaper across that suite than Gemini 3.5 Flash-Lite running through the Interactions API. The written announcement does not list the underlying scores.
Several Google products already use it. Gemini Notebook routes queries between models with the API and saw up to 40% lower latency on everyday requests, according to Google. Stitch uses it to generate playable wireframes and apply voice or tap edits, CC from Google Labs uses it to pick an emoji reaction from more than 200 options, and Google Search uses it as a safety layer for MCP tools.
The endpoint sits on the Gemini Developer API's v1beta channel and requires version 3.2 or later of the google-genai SDK. Google's documentation lists some early limits: no context caching, Batch API or Flex and Priority inference, inline uploads capped at 100 MB per request (50 MB for PDFs), and no support for YouTube links or Cloud Storage URIs.
Integrations are rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode. Google said support for the Decisions API in Gemini in Google Cloud is coming soon.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
