Google introduces gemini decisions API for closed-set scoring at $0.10/Million tokens

Google has released the Gemini Decisions API, featuring the specialized model Gemini Decision 1 (gemini-decision-1). This API is designed not for text generation, but for scoring up to 1,000 questions against a single input, returning calibrated probabilities. It is intended for use in agent workflows for tasks like routing and tool selection. The pricing structure is set at $0.10 per million input tokens, with no charge for output tokens. Gemini Decision 1 processes multimodal inputs—including text, images, video, audio, and PDFs—in one pass, with zero output token usage. Benchmarks self-reported by Google show Gemini Decision 1 achieving an error rate of 0.0250, which is roughly half that of TypeSafe Jev 1.14's 0.0494. Furthermore, Google claims the model is nearly five times cheaper than Flash-Lite across the shared suite, costing $0.88 versus $4.14. The endpoint is currently available on the Gemini Developer API (v1beta), with a no-cost tier accessible via Google AI Studio for prototyping. Ecosystem integration is expanding across platforms such as LangChain and OpenRouter, while Google internally uses the model in applications like Gemini Notebook and Google Labs’ CC.