Google launches Gemini Decision 1 for the small choices inside AI agents
Google has introduced Gemini Decision 1 and a dedicated Decisions API for bounded tasks such as choosing a tool, routing a request or grading retrieved context. Its November 11 API changelog confirms the release through the Gemini API.
According to Google’s announcement, developers can submit text, images, audio, video or PDFs with up to 1,000 questions against a shared input. The model scores yes/no conditions, candidate choices and ordered rubrics rather than composing a conventional answer. Google lists a 256K-token input window and pricing of $0.10 per million input tokens, with no output-token charge. Shared input is billed once per request; question instructions and criteria still add tokens.
Google reports 150–250 millisecond end-to-end latency in tests covering contexts up to 16K tokens and one to 100 questions. That is not a guarantee at the maximum input size or question count. We have not tested access or reproduced the performance claims.
The practical significance is the separation of a workflow’s small decisions from its expensive generative work. A low-cost classifier could decide which model or tool to invoke before a larger system starts writing or acting. But confidence scores are not proof of correctness, and a probabilistic policy check should not be confused with a hard permission boundary.
Sources and citations
Revision history
Published .