Gemini 4 Hydrogen
A multimodal developer model for everyday coding, agent workflows and knowledge work with configurable thinking levels.
Gemini Decision 1
aistudio.google.com
Multimodal decision model returns calibrated probabilities for binary, categorical, and rubric questions without generating output tokens.
Visit website
Pricing
Freemium
Available on
Gemini Developer API, Google AI Studio
Category
Llm Models
What is Gemini Decision 1?
Gemini Decision 1 is Google’s multimodal model for bounded judgments such as classification, binary checks and rubric scoring. Accessed through the Decisions API, it returns calibrated probabilities without generating text tokens.
Google positions it for developers building tool selection, policy checking, model routing and context grading into applications. It is suited to workflows that need a structured judgment to guide the next action. The model identifier is gemini-decision-1.
Key features
Multimodal evaluation
Evaluate shared text, PDFs, images, audio and video.
Typed questions
Define yes/no checks, categorical choices or rubric scores.
Independent questions over shared input
Submit up to 1,000 questions within a shared 262,144-token context. Each question evaluates the shared input independently and cannot see other questions; compose dependent logic in application code.
Application routing
Use probability-based decisions for tool selection, model routing or context grading.
Who is it for?
Developers designing request routing
Evaluate predefined categories to decide which application workflow should handle an incoming request.
Teams applying review rubrics
Define scoring criteria for assessing submitted material before a human review.
Agent builders selecting tools
Assess which available tool best matches a task using predefined choices.
How to use Gemini Decision 1
- Choose a Python, JavaScript or REST example from the Decisions API documentation.
- Configure API authentication and select gemini-decision-1.
- Provide shared input and define typed questions with the required choices or criteria.
- Submit the request and combine the returned probabilities in your application logic.
Gemini Decision 1 pricing
Freemium
A free tier is available alongside Standard paid usage. Shared input is billed once per request; question instructions, criteria and formatting add input tokens. No output tokens are billed.
Standard paid tier
USD 0.10 per million input tokens
Usage-based API billing, not a monthly subscription
Applies to text, image, video, audio and PDF input. The model generates zero output tokens.
What to know before you use it
- Access is limited to Gemini Developer API v1beta with Gemini Decision 1.
- Question instructions accept text only. YouTube URLs and direct gs:// media URIs are unsupported.
- Yes/no probabilities range from 0.01 to 0.99; choice questions and scoring rubrics permit 2–255 options or levels.
- Batch API and context caching are unsupported. Free-tier content is used to improve Google’s products; paid-tier content is not, subject to service terms.
Frequently asked questions
Does it produce written explanations?
No. The Decisions API returns typed probability answers and generates no text tokens.
Is shared input charged separately for every question?
No. Shared input is billed once per request, while each question contributes its own instruction, criteria and formatting tokens.
Where does it fit in an agent workflow?
Google lists tool selection, policy checking, model routing and context grading as intended applications.