Introduction to Gemini Decision 1

Gemini Decision 1 is Google's purpose-built decision model behind the Gemini Decisions API. Its model ID is gemini-decision-1. It is designed for bounded judgment tasks: an application supplies shared input and a set of questions, and the model scores candidate answers and returns probabilities and confidence rather than composing a long natural-language response token by token.

A single request can use shared multimodal input—text, images, audio, video, and PDFs—and evaluate up to 1,000 independent questions. The model is accessed through the Gemini API in Google AI Studio.

Key Features

• Three structured question types: noul for yes/no probability, choice for selecting from fixed options, and score for grading against an ordered rubric.
• Multimodal input: Process text, images, audio, video, and PDFs in a shared input.
• Batch evaluation: Evaluate up to 1,000 independent questions in one request against the same input.
• Probabilities and confidence: Receive answer probabilities or scores plus a confidence value from 0 to 1, helping applications decide whether to automate or escalate a result.
• Low-latency design: Google reports 150–250 ms end-to-end latency; actual performance depends on the input and calling environment.
• API and SDK support: Official guides include Python, JavaScript, and REST examples using the Google GenAI SDK.

What Makes Gemini Decision 1 Different?

Traditional generative models produce text token by token, which an application may then need to parse. Gemini Decision 1 is built for closed-set decisions: it scores predefined questions and candidate options without generating answer text tokens. This can reduce parsing overhead and make it practical to evaluate many independent checks against one shared input.

Each question is evaluated independently against that input; questions do not see one another's answers. The intended pattern is to ask narrow, single-purpose questions and combine their results in application code, rather than relying on one broad prompt to perform an entire workflow.

How to Use Gemini Decision 1

1. Create an API key in Google AI Studio.
2. Install the google-genai SDK (the documentation specifies version 3.2 or later), or use the JavaScript SDK or REST interface.
3. Call the Decisions API with model gemini-decision-1 and provide the shared input.
4. Define each question with a type and clear instructions: noul, choice, or score.
5. Read the returned answer, probabilities, and confidence. Set thresholds for low-confidence results and route them to a fallback model or human reviewer.

Use Cases

• Model routing: Choose a model or processing path based on request type or complexity.
• Tool selection: Select the most appropriate tool or action from a predefined set.
• Moderation and policy checks: Test input against explicit rules.
• RAG relevance evaluation: Score whether retrieved material is relevant to a query.
• Batch media review: Run multiple independent checks over image, audio, video, or PDF content.
• Operations and quality review: Grade tickets, records, or content against a fixed rubric.

Who Is Gemini Decision 1 For?

• Developers building AI agents, model routers, and automated workflows.
• Product and platform teams running moderation, policy checks, or classification at scale.
• Engineering teams evaluating multimodal material for quality, ranking, or relevance.
• API users who want to reduce the overhead of generating and parsing text for simple decisions.

Pricing and Availability

Google lists pricing at $0.10 per 1 million input tokens, with no charge for output tokens. The shared input is billed once per request; question instructions and criteria also contribute to input-token usage. Google AI Studio offers an API no-cost tier for prototyping, subject to the current quotas and access conditions shown in the console and official documentation.

The model is available through the Gemini API in Google AI Studio. Google's announcement links to Python, JavaScript, and REST guides as well as example apps. Confirm the current API release status, quotas, and regional availability before deploying it.

Pros and Cons

Frequently Asked Questions (FAQ)

Is Gemini Decision 1 a general-purpose chatbot?

No. It is a specialized model for closed-set decisions. It returns structured answers, probabilities, and confidence rather than general long-form conversation.

Which input types does it support?

Google lists text, images, audio, video, and PDFs, which can be combined in one shared input.

How many questions can one request evaluate?

Up to 1,000 independent questions can be evaluated against the same shared input.

Which question types are available?

The API supports noul (yes/no), choice (fixed options), and score (ordered rubric scoring).

How is Gemini Decision 1 priced?

Google lists $0.10 per million input tokens and no output-token charge. Google AI Studio also offers an API no-cost tier for prototyping; current quotas and conditions apply.

Does it replace Gemini's generative models?

Not directly. It is suited to fast, structured judgments. Use a generative model when a task requires explanations, creative output, or open-ended answers.