Google launches a Gemini API that answers in probabilities

Developers building apps or AI agents get a cheaper way to make small decisions, if Google's pricing holds. Gemini Decisions API costs $0.10 per million input tokens and charges nothing for output. Google unveiled it Nov. 11 inside the Gemini API and Google AI Studio, with a free tier for prototypes. Ordinary users see no direct change, and any benefit would come indirectly, if apps run faster or cost less to operate. Google has given no price or availability specific to Mexico.

How it differs

The service runs on a new model, Gemini Decision 1, that does not write answers. It takes an input (text, image, audio, video or PDF), checks it against a fixed list of questions and returns a probability for each.

One request can carry up to 1,000 questions about the same input, with a context window of up to 256,000 tokens, the units models use to count text. Questions come in three formats: yes or no, a choice among 2 to 255 options, and an ordered rubric that yields a weighted score. The shared input is billed once, and each question adds only its own tokens.

Google pitches it for routing queries, picking tools, safety checks and grading content. The company says it already uses the model in Gemini Notebook, to decide which model handles each query, with up to 40% lower latency on everyday queries, and in Search, as a safety layer for MCP tools, the protocol agents use to connect to outside services.

What hasn't been tested

Every figure comes from Google, and there is no independent testing. The company reports 50 to 100 milliseconds of model compute and 150 to 250 end to end with contexts of up to 16,000 tokens. It also says that, on its own test set, the service costs a little more than a fifth of what Gemini 3.5 Flash-Lite costs through the Interactions API. Its calibration charts compare the model with "other APIs" without naming them, and latency is reported as stable only from 1 to 100 questions, not at 1,000.

Announced integrations include OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode. Gemini in Google Cloud is listed as "coming soon," with no date or regions.