Google announced Gemini Decision 1 for fast closed-set answers

Google DeepMind announced the Gemini Decisions API, a new endpoint powered by a model it calls Gemini Decision 1, which it says is built for fast, closed-set decisions rather than long written answers. The company blog post, by Min Choi and Felix Yu, is dated November 11, 2026, with a publication timestamp of 6:00 p.m. UTC, which is 1:00 PM ET. This article summarizes that announcement and contemporaneous coverage by SiliconANGLE. It is not a hands-on test of the API, and the speed, cost, and accuracy figures below are company claims unless a separate outlet is quoted.

Real, reported: based on the public sources cited below. Nothing added or dramatized. Sources checked November 13, 2026, at 7:14 AM ET (EST). Latency, pricing, and benchmark statements come from Google's post or from SiliconANGLE's account of that post. The opened sources do not present this release as a peer-reviewed paper. Those results should be read as a company claim, not as an independently verified benchmark.

What Google says the API does

The post says: "Today, we're introducing the Gemini Decisions API, a new endpoint for fast, closed-set decisions powered by our new Gemini Decision 1 model (gemini-decision-1), available through the Gemini API in Google AI Studio." It describes the product as providing "purpose-built primitives for bounded choices like tool selection, policy checks, model routing, and context grading."

In Google's description, a developer can "pass one multimodal input across text, images, audio, video, and PDFs (up to 256K tokens) along with up to 1,000 independent questions." Instead of "generating answer text token by token," the post says, "Gemini Decision 1 scores your candidate options in a single forward pass, returning probabilities and confidence scores in 150-250 ms end-to-end latency at $0.10 per 1M input tokens with zero output tokens billed."

The blog names three question types that can be mixed in one request: "noul (Yes/No)," which "estimates the probability that a yes/no condition holds"; "choice (Multiple Choice)," which "selects from 2 to 255 choices, returning the winner and normalized probabilities for every candidate"; and "score (Ordered Rubric)," which "grades inputs across an ordered rubric (starting at 0), returning a continuous weighted score and per-level probabilities." "Every answer includes a confidence score (from 0 to 1) so your code can route ambiguous cases to a fallback or human reviewer."

On pricing, Google said Gemini Decision 1 "is priced at $0.10 per 1M input tokens and $0 for output tokens, with an API no-cost tier you can use in Google AI Studio for prototyping." It also said the shared input "is billed once per request no matter how many questions you evaluate (each question adds only its own instruction and criteria tokens)," and that there are "no output-token charges, variable thinking-token bills, or separate audio transcription and OCR costs."

Performance and product claims in the official post

Under the heading "Strong performance and calibration," the post states: "On benchmarks spanning agent routing, reasoning, ranking, and multimodal understanding, Gemini Decision 1 leads across the suite, with the lowest calibration error." It does not, in the passages we cite, publish a full table of those benchmarks or name an outside auditor. That "leads across the suite" line is a company claim.

On speed, Choi and Yu wrote: "Gemini Decision 1 scores options without token-by-token decoding, keeping median latency 50-100 ms of model compute (150-250 ms end-to-end), across multimodal inputs and contexts up to 16K tokens, remaining nearly flat from 1 to 100 questions per request." The post also says that, "with zero output tokens billed, Gemini Decision 1 with the Decisions API is nearly 5x cheaper across the shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API." The "nearly 5x cheaper" comparison is Google's, for the shared suite it describes, not a price we independently calculated.

The announcement lists internal uses, which Google presents as examples rather than as customer case studies we verified: Gemini Notebook using the API "as a real-time router"; a rapid-canvas experiment in Stitch; emoji selection in a Google Labs product called CC; and a "safety layer for MCP tools in Search" that, Google says, detects "adversarial behavior and content against adversarial evals with high efficacy within the latency envelope." Those are the company's descriptions of its own products.

Google said developers can start in Google AI Studio, and that "ecosystem integrations are rolling out across OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit, and OpenCode alongside the google-genai SDKs, with support coming soon to Gemini in Google Cloud."

How SiliconANGLE described the same launch

SiliconANGLE, in an article by Maria Deutscher first posted at 8:22 PM ET on November 11 (the opened page also shows a later update), reported that Google "today launched Gemini Decisions API, an artificial intelligence service designed to answer user questions in a cost-efficient manner," powered by Gemini Decision 1. The outlet wrote that the model "can process just three types of questions" — yes/no, multiple choice, and a user-specified scale — and that, "according to Google, the model is nearly five times cheaper than its entry-level Gemini 3.5 Flash-Lite LLM." That cost comparison is SiliconANGLE's paraphrase of Google's "nearly 5x cheaper" claim; the official post ties the comparison to "the shared benchmark suite" and to Gemini 3.5 Flash-Lite used with the Interactions API.

SiliconANGLE also quoted the official latency sentence directly: "Gemini Decision 1 scores options without token-by-token decoding, keeping median latency 50-100 ms of model compute (150-250 ms end-to-end), across multimodal inputs and contexts up to 16K tokens, remaining nearly flat from 1 to 100 questions per request." It correctly reported the 256,000-token multimodal input window, the 10 cents per 1 million input tokens price, and the fact that Google "doesn't charge for output tokens," plus a limited free tier in Google AI Studio.

The SiliconANGLE article also places the launch in a wider "decision model" market and reports additional competitive comparisons. Where those comparisons do not appear in the official passages we cite, they remain SiliconANGLE's account and are not repeated here as Google's own published scores.

How to read the claims

Google is announcing a developer API and a specialized model, not an independently audited benchmark suite in the materials we opened. Latency ranges, the "nearly 5x cheaper" comparison, and the claim that the model "leads across the suite, with the lowest calibration error" should be read as company results. A confidence score returned by the model is still the model's own estimate, not a guarantee that the answer is right. The official post itself says the score is there so code can "route ambiguous cases to a fallback or human reviewer."

Our advice (general guidance)

If you build on this API, treat yes/no, routing, and scoring outputs as machine estimates that need a threshold and a review path, especially for safety, billing, or account changes. Check the official documentation and pricing in Google AI Studio or the Gemini API docs you already use, rather than through an unofficial recap. Company latency and cost comparisons are not a substitute for measuring the API on your own traffic.

These are editorial precautions, not individualized legal, financial or security advice.

Sources

Published November 13, 2026 by FraudMeNot Team. Label: Real, reported.

Disclaimer. General education only, not legal, financial or security advice. No guarantee. Consult a qualified professional. News summaries are based on the sources listed and may change as more is reported.

General education only, not legal, financial or security advice. Not affiliated with the organizations discussed. No sponsorship or affiliate links. Information may change as further evidence becomes available.