Google has introduced the Gemini Decisions API, a new endpoint built for bounded choices such as tool selection, policy checks, model routing and grading context. It runs on a new model, Gemini Decision 1 (gemini-decision-1), and is available through the Gemini API in Google AI Studio. The announcement is dated Nov 11, 2026.
What it does
The model does not write an answer. It scores the candidate options you supply in one forward pass and returns probabilities and confidence scores. You send a single input, which can combine text, images, audio, video and PDFs up to 256K tokens, plus as many as 1,000 independent questions about it.
There are three question types:
- Yes/no: the probability that a condition holds, with optional descriptions of what counts as true or false.
- Choice: a pick from 2 to 255 options, with normalized probabilities for every candidate.
- Score: a grade against an ordered rubric, returning a weighted score and a probability per level.
Every answer carries a confidence value between 0 and 1. Google's stated intent is that your own code uses it to send ambiguous cases to a fallback or a human reviewer.
The examples in the post show modalities mixed in one request: a call recording alongside account metadata, or a scanned PDF claim alongside damage photos. Resolution on image and video inputs can be tuned to control detail and token cost.
Speed and price
Google reports 150-250 ms end-to-end latency, with 50-100 ms of that being model compute, and says latency stays nearly flat from 1 to 100 questions per request. That latency figure is stated for contexts up to 16K tokens. Pricing is $0.10 per 1M input tokens, and output tokens are billed at zero. The shared input is charged once per request, and each added question costs only its own instruction and criteria tokens. There are no separate charges for audio transcription or OCR, and no thinking-token bills.
Google also claims the combination is nearly 5x cheaper across its shared benchmark suite than Gemini 3.5 Flash-Lite used through the Interactions API, and that Gemini Decision 1 has the lowest calibration error on that suite. Both are Google's own benchmark results, not independent tests.
Early use inside Google
Google describes several internal uses. Gemini Notebook routes incoming queries to a model tier and thinking level, which Google says cuts latency by up to 40% on everyday requests. Google Search uses the API as a safety layer to detect adversarial behavior around MCP tools. Other examples are design prototyping in Stitch and emoji selection in a Labs product.
Availability
A no-cost API tier in Google AI Studio is available for prototyping, with guides for Python, JavaScript and REST. Integrations are rolling out with OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode, next to the google-genai SDKs. Support in Gemini in Google Cloud is listed as coming soon, with no date given.
Top comments (0)