Gemini Decisions API

Name: Gemini Decisions API
Author: Google DeepMind

Gemini Decisions API is a new Gemini API endpoint, powered by the gemini-decision-1 model, that scores yes/no, multiple-choice and rubric questions (up to 1,000 per request) against one text, image, audio, video or PDF input in a single pass and returns probabilities and confidence scores instead of generated text.

Visit Gemini Decisions API ↗

Paid (usage-based), free tierProprietary (hosted API)REST API; Python and JavaScript SDKs (google-genai)by Google DeepMind

Why it matters

For developers building agents and apps that make lots of small, bounded calls (tool selection, model routing, policy checks, grading) who want fast, cheap, calibrated answers without training custom classifiers or paying for generated tokens.

Highlights

Three question types: yes/no probability, choice among 2 to 255 options with normalized probabilities, and ordered-rubric scores; every answer includes a 0-1 confidence score for routing unclear cases to a fallback or human.
One shared multimodal input (text, images, audio, video, PDFs, up to 256K tokens) can be checked against up to 1,000 independent questions in one request.
Google reports 150-250 ms end-to-end latency (50-100 ms of model compute) that stays nearly flat from 1 to 100 questions, and says it is nearly 5x cheaper on its benchmark suite than Gemini 3.5 Flash-Lite via the Interactions API.
Google says it already powers model routing in Gemini Notebook, rapid canvas edits in Stitch, emoji picks in CC and an MCP-tool safety layer in Search.
Integrations are rolling out for OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode alongside the google-genai SDKs.

Taken from the project's own README, docs or launch page. Claims about speed or accuracy are the maker's unless stated otherwise.

Try it if: Your agent spends a full LLM call just to pick a tool, route a request or check a policy, and you want that decision in a few hundred milliseconds with a confidence score.

Worth knowing: Benchmark, latency and cost comparisons are Google's own figures. It answers closed-set questions only; it doesn't generate text.

Quick facts

Pricing	$0.10 per 1M input tokens, $0 for output tokens; the shared input is billed once per request; no-cost tier for prototyping in Google AI Studio
Free or paid	Paid (usage-based), free tier
License	Proprietary (hosted API)
Language	REST API; Python and JavaScript SDKs (google-genai)
Platform	Gemini API via Google AI Studio; Google Cloud support announced as coming soon
GitHub stars	n/a (not a GitHub project)
Maker	Google DeepMind

Related ranking

See how established tools compare in Best AI Coding Tools for Developers. New releases aren't scored until we review them against the rubric.

In the news

Google launches the Gemini Decisions API for fast, low-cost yes/no, choice and rubric scoring November 11, 2026

Checked 2026-11-12 from the model's own pages. Prices and star counts change, so confirm on the official site. Listing here is not a ranking or an endorsement.