Start your day with intelligence. Get The OODA Daily Pulse.

Home > Briefs > Technology > Google Launches Gemini Decisions API and Specialized “Gemini Decision 1” Model for High-Speed Multimodal Classification

Google Launches Gemini Decisions API and Specialized “Gemini Decision 1” Model for High-Speed Multimodal Classification

Google officially launched the Gemini Decisions API, a specialized developer endpoint designed for rapid, high-throughput closed-set classification and decision-making against unstructured data. The service is backed by a dedicated underlying model, Gemini Decision 1, engineered explicitly to evaluate structured questions (yes/no, multiple choice, and rubric-based checks) rather than generating open-ended conversational text. Developers can evaluate up to 1,000 typed questions in a single pass against a shared multimodal input (documents, images, or audio), returning calibrated probabilities with zero text-token overhead. Priced aggressively (around $0.10 per 1M input tokens), the API targets high-volume enterprise workflows such as customer support classification, automated compliance checks, and document auditing by trading generative token bloat for raw, structured decision telemetry. The release arrives alongside Google’s concurrent rollout of agentic execution frameworks (such as Gemini 4 Carbon), underscoring an industry-wide race to optimize developer tools for specialized agent pipelines and high-volume data orchestration.

Full report : New developer endpoint enables simultaneous evaluation of up to 1,000 closed-set or rubric questions per call with zero text-token generation, dramatically cutting cost and latency for enterprise document processing.