Key Notes
- Gemini Decision 1 evaluates predefined choices and returns structured probability and confidence information for applications and AI agents.
- Google quotes 150–250 ms latency for contexts up to 16,000 tokens, while the complete request can contain up to 262,144 tokens.
- The service costs $0.10 per million input tokens, generates no text output tokens and includes a free tier for testing.
Google has launched Gemini Decision 1, a specialized model that chooses between predefined answers instead of writing a conversational response. The company introduced it on November 11 through the Gemini Decisions API, targeting the small, repeated judgments that determine what an application or AI agent does next.
The service accepts text, images, audio, video and PDFs. A support application, for example, could examine a complaint and a product photograph, decide which department should receive the case and assess whether it needs human attention. The response contains structured scores that software can use directly.
Up to 1,000 Questions About the Same Input
Google’s developer guide specifies three question types: yes-or-no checks, selection from a list and scoring against an ordered rubric. One request can contain up to 1,000 independent questions, with up to 255 options for a multiple-choice question. Every question is evaluated independently against the shared evidence and cannot read the answers to other questions.
Google reports end-to-end latency of 150–250 milliseconds for contexts up to 16,000 tokens, with latency remaining nearly flat from one to 100 questions in its testing. That measured range should not be read as a guarantee for the maximum request. The documented context ceiling is 262,144 tokens across the shared input and every question, including instructions and answer criteria.
Input Costs $0.10 per Million Tokens
The published pricing is $0.10 per million input tokens across supported media. The model generates no text tokens, so there is no output-token bill. A free tier lets developers test the service through Google AI Studio.
Shared material is billed once within each request, regardless of how many questions examine it. Each question still adds its own instruction and criteria tokens. Asking several checks about one document therefore avoids repeatedly charging for that same document inside the call. Context caching is not supported.
Google lists free-tier quotas of 200 requests per minute, 50,000 input tokens per minute and 10,000 requests per day. Those limits apply together; a developer can exhaust the token allowance before reaching the request ceiling.
Another Contender for the Decisions Inside AI Agents
The launch joins a category AIstify has followed as Microsoft launched Decision-1 and OpenAI introduced its Decisions API. These services make selection among allowed outcomes a distinct part of the AI stack, alongside the larger models that plan tasks or generate content.
That division still leaves developers responsible for defining sensible choices. Google’s guide notes that a multiple-choice question distributes probabilities across the supplied options even when none fits. An explicit fallback option and confidence thresholds can send ambiguous cases to a reviewer instead of forcing the application down the wrong path.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
