Gemini Decisions API: Cheap, fast yes/no and routing calls

Written by , Founder of Launch Llama · November 12, 2026

TL;DR
  • The Gemini Decisions API, powered by the new gemini-decision-1 model, scores yes/no, multiple choice and rubric questions in a single forward pass instead of generating text.
  • Google lists 150-250 ms end-to-end latency and $0.10 per 1M input tokens, with $0 for output tokens. There is a no-cost tier in Google AI Studio for prototyping.
  • One input (text, images, audio, video, PDFs, up to 256K tokens) can be checked against up to 1,000 questions, each returning a 0 to 1 confidence score.

What happened

On Nov 11, 2026, Google introduced the Gemini Decisions API, a new endpoint for fast, closed-set decisions. It runs on a new model, Gemini Decision 1 (gemini-decision-1), and is available through the Gemini API in Google AI Studio.

Instead of writing an answer token by token, the model scores your candidate options in a single forward pass and returns probabilities and confidence scores. You pass one multimodal input along with up to 1,000 independent questions. There are three question types: yes/no, choice (2 to 255 options) and score (an ordered rubric starting at 0).

Google says median model compute is 50-100 ms (150-250 ms end-to-end) and stays nearly flat from 1 to 100 questions per request, measured on contexts up to 16K tokens. It also says the API is nearly 5x cheaper across its shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API. That is Google's own benchmark claim, and the source gives no further detail on it.

What this means for founders

  • Cost is simple to predict. Pricing is $0.10 per 1M input tokens and $0 for output. The shared input is billed once per request, and each extra question adds only its own instruction and criteria tokens. Google says there are no thinking-token bills or separate transcription and OCR costs.
  • Good fit for routing and gating. Google names tool selection, policy checks, model routing and context grading as use cases. Gemini Notebook uses it to route queries to a model tier, which Google says gives up to 40% lower latency for everyday queries.
  • Confidence scores let you add a safety net. Every answer has a 0 to 1 score, so your code can send ambiguous cases to a fallback or a human reviewer instead of trusting a guess.
  • Less custom work. Google says teams ship features by defining options rather than training custom classifiers. Whether that holds for your data is something to test.
  • Risks to check. The speed, cost and calibration claims come from Google. The code sample uses the v1beta API version, so expect possible changes. Support for Gemini in Google Cloud is listed as coming soon.

What the Launch Llama directory shows

Live numbers from Launch Llama for listings that describe themselves as "ai decision engine", "llm api", "ai workflow automation", "decision automation api" tools, as of 2026-11-12.

SignalLaunch Llama data
Listings in this space5
New in the last 30 days1
With a free or freemium plan67%
Most upvoted by foundersEachAPI (21 upvotes)

Launch Llama has 5 matching listings in this space, 1 of which went live in the last 30 days, and 67% have a free or freemium plan. The most upvoted is EachAPI with 21 upvotes.

How to try it today

  1. Get an API key in Google AI Studio. Google says there is a no-cost tier for prototyping.
  2. Try the demo apps Google lists: Doorstep Parcels, Speed Sketch and Find the Zoo Keeper.
  3. Follow the Python, JavaScript or REST guide in the Decisions API documentation.
  4. Pick one narrow decision in your product, such as a yes/no policy check or routing a request to one of a few options, and define it as a question with criteria.
  5. Use the confidence score to decide when to fall back to a human or a bigger model.
from google import genai

client = genai.Client(http_options={"api_version": "v1beta"})

result = client.decisions.create(
    model="gemini-decision-1",
    input=[{"type": "text", "text": "Your input here"}],
    questions={
        "has_spill": {
            "type": "noul",
            "instructions": "Is there an uncleaned liquid spill on the counter?",
        },
    },
)

This is adapted from Google's example, which uses "noul" as the yes/no question type. Check the docs for the current type names.

Related tools on Launch Llama

Tools from the Launch Llama directory that fit with this news:

ToolWhat it doesWhy it matters here
Gemini 3.1 Flash-LiteFor AI engineers: Gemini 3.1 Flash-Lite is a lightweight model for high-volume, latency-sensitive agent pipelines. Handles tool calling, classification, and multimodal processing.Gemini 3.1 Flash-Lite is a lightweight model for classification and latency-sensitive agent pipelines, a job the Decisions API now targets with a purpose-built endpoint.
Lightning Rod: AI Forecasting APIFor developers: Lightning Rod: AI Forecasting API scores calibrated predictions on future events so your agents and bots make data-driven decisions faster.Lightning Rod also focuses on calibrated scores for agents making decisions, though it covers predictions on future events rather than closed-set choices over your own inputs.
One API for the World’s Leading AI ModelsFor developers: API-Route provides one OpenAI-compatible gateway to GPT, Claude, Gemini, and other models so you switch AI providers without code changes.API-Route is an OpenAI-compatible gateway covering Gemini and other models; whether it supports the Decisions API is not confirmed.
ApiariumFor developers: Apiarium unifies OpenAI, Claude, and Gemini APIs so you switch models without rewriting code.Apiarium unifies OpenAI, Claude and Gemini APIs; Decisions API support is not confirmed, but it is relevant if you want to avoid lock-in to one provider.

Browse all tools or list your own tool.

Llama's take

I would try this now if you run a product with a lot of small, repeated judgment calls: routing requests, checking content against rules, grading inputs. Paying only for input tokens and getting a confidence score back is a cleaner setup than parsing generated text. The free tier in AI Studio makes a test cheap.

If your decisions are open-ended or need explained reasoning, wait. This API only picks from options you define. If you need a provider-neutral stack, also wait until the integrations with OpenRouter, LangChain, Vercel, LiteLLM and others are live. Google says they are rolling out, but I can't confirm which are available today.

The thing to watch is calibration on your own data. Google says Gemini Decision 1 has the lowest calibration error on its benchmarks, but that is Google's own suite. Run your real cases through it before you trust the confidence scores.

FAQ

What is the Gemini Decisions API?

It is a new endpoint from Google, powered by the gemini-decision-1 model, for fast closed-set decisions. It scores yes/no, multiple choice and rubric questions in a single forward pass and returns probabilities and confidence scores instead of generated text.

How much does the Gemini Decisions API cost?

Gemini Decision 1 is priced at $0.10 per 1M input tokens and $0 for output tokens. The shared input is billed once per request, and each question adds only its own instruction and criteria tokens. There is a no-cost API tier in Google AI Studio for prototyping.

What inputs does the Gemini Decisions API accept?

It accepts text, images, audio, video and PDFs, up to 256K tokens, in one shared input. You can attach up to 1,000 independent questions to it. Choice questions support 2 to 255 options, and every answer includes a confidence score from 0 to 1.

Sources

Related guides

Explore founder tools

Free calculators, plus newsletters and launch directories curated for founders.

Mini Apps →Newsletters →Launch Directories →

For bots & AI agents

Machine-readable versions of this article and Launch Llama. Fetch these instead of parsing this page.

This article (.md) llms.txt llms-full.txt ai.txt Full catalog (.md) Endpoint index API docs MCP server