# Gemini Decisions API: Features, Pricing and How to Use It

By Schnalz.net · 12 November 2026

https://schnalz.net/ai-guides/gemini-decisions-api-features-pricing-and-how-to-use-it

> The Gemini Decisions API is a tool from Google DeepMind that scores up to 1,000 closed-set questions across multimodal inputs in a single forward pass. Powered by the Gemini Decision 1 model, it returns confidence scores with low latency and zero output token fees.

## In short

- Google DeepMind released the Gemini Decisions API powered by the Gemini Decision 1 model.
- It processes text, images, audio, video, and PDFs in a single pass for up to 1,000 questions.
- Input costs $0.10 per million tokens with no charge for output tokens.
- It supports yes/no checks, multiple choice options, and rubric scores with calibrated confidence levels.

## What is the Gemini Decisions API?

The Gemini Decisions API is an endpoint launched by Google DeepMind on 11 November 2026 to execute fast, closed-set evaluations without token-by-token text generation. Powered by the Gemini Decision 1 model (gemini-decision-1), it allows developers to score bounded choices such as tool selection, content grading, and policy rules.

**Gemini Decisions API at a glance**

- **Developer**: Google DeepMind
- **Model ID**: gemini-decision-1
- **Launch date**: 11 November 2026
- **Input limit**: Up to 256K tokens
- **Input price**: $0.10 per 1M tokens
- **Output price**: $0 per token
- **Latency**: 150–250 ms end-to-end

The Gemini Decisions API evaluates up to 1,000 independent questions in a single request against a shared input context. It accepts native multimodal inputs combining text, images, audio, video, and PDFs in one request.

## How much does the Gemini Decisions API cost?

The Gemini Decisions API costs $0.10 per 1 million input tokens with zero charges for output tokens. Google AI Studio also provides a no-cost tier for developers prototyping with the Gemini Decision 1 model.

Because Gemini Decision 1 scores candidate choices directly in a single forward pass instead of generating text tokens, users are not billed for output tokens, variable thinking tokens, or separate OCR and audio transcription fees. Shared inputs are billed only once per request regardless of how many questions are evaluated, with each question adding only its own instruction tokens.

## What question types does the Gemini Decisions API support?

The Gemini Decisions API supports three distinct primitive question types for bounded decision-making. Developers can mix these question types within a single request against shared multimodal content.

Every response includes a confidence score between 0 and 1 so applications can route uncertain cases to human reviewers or fallback models. The available question primitives include:

- noul (Yes/No): Evaluates the probability that a condition holds, with optional criteria for true and false states.
- choice (Multiple Choice): Selects the winning option from 2 to 255 choices and returns normalized probabilities for all candidates.
- score (Ordered Rubric): Grades inputs against an ordered rubric starting at 0, returning per-level probabilities and a continuous weighted score.

## Gemini Decisions API vs Competitors

The Gemini Decisions API competes with decision model endpoints like Jev and OpenAI's GPT-6 Luna model. While Jev and OpenAI limit input types, Gemini Decision 1 supports full native multimodal inputs across text, images, audio, video, and PDFs.

**Decision model comparison**

| Model | Input price (per 1M tokens) | Output token cost | Supported modalities |
| --- | --- | --- | --- |
| Gemini Decision 1 | $0.10 | $0 | Text, images, audio, video, PDFs |
| GPT-6 Luna | $0.10 | $0 | Text, images |
| Jev | $0.042 | Not stated | Text and JSON only |

According to weblog tests published by Simon Willison, Gemini Decision 1 returns continuous probability scores (such as 0.05 or 0.92) for yes/no questions rather than absolute 1.0 or 0.0 values.

## How to use the Gemini Decisions API

Developers can access the Gemini Decisions API by obtaining an API key in Google AI Studio and using the official Python SDK or REST endpoints. Google DeepMind provides SDK support through the google-genai library.

1. Obtain an API key in Google AI Studio.
2. Initialize the Google GenAI client in Python specifying API version v1beta.
3. Call client.decisions.create with model set to gemini-decision-1, passing your multimodal input blocks and a questions dictionary.
4. Inspect the returned probabilities and confidence scores to make programmatic routing or grading decisions in your application.

> **Tip**: You can control token costs and detail level by tuning the image and video resolution settings on input files sent to the Gemini Decisions API.

## What are the limits of the Gemini Decisions API?

The Gemini Decisions API is designed for bounded, closed-set choices rather than open-ended text generation or long-form output. It does not return decoded text answers, explanatory reasoning text, or generated content.

Each request is capped at 256K input tokens and up to 1,000 questions. In addition, yes/no queries return probabilistic floating-point estimates rather than hard binary outputs.

## Frequently asked questions

### Who developed the Gemini Decisions API?

Google DeepMind developed the Gemini Decisions API, introducing it on 11 November 2026 alongside the Gemini Decision 1 model.

### What input formats does the Gemini Decisions API accept?

The Gemini Decisions API accepts text, images, audio, video, and PDF files. Developers can combine multiple modalities into a single shared input up to 256K tokens.

### How is the Gemini Decisions API priced?

The Gemini Decisions API costs $0.10 per 1 million input tokens and charges $0 for output tokens. A free tier is available in Google AI Studio for testing and prototyping.

### Where can developers access the Gemini Decisions API?

Developers can access the Gemini Decisions API in Google AI Studio by creating an API key. Ecosystem support is also rolling out to tools like OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, and LiveKit.

### How does the Gemini Decisions API achieve low latency?

The Gemini Decisions API achieves low latency by scoring candidates in a single forward pass instead of using token-by-token decoding. End-to-end latency ranges from 150 to 250 milliseconds.

## Sources

- [Google: Gemini Decisions API: Fast multimodal decisions](https://blog.google/innovation-and-ai/technology/developers-tools/gemini-decisions-api/)
- [Simon Willison: Gemini Decisions API](https://simonwillison.net/2026/Nov/11/gemini-decisions-api/)