Google has launched Gemini Decision 1, a model and API built for closed-set decisions. Rather than generating text token by token, it scores the available choices in a single pass. Google said the model supports yes/no, multiple-choice (up to 255 options) and rubric scoring. It accepts text, image, audio, video and PDF inputs, handles up to 1,000 questions per request within a 256K context window, and attaches a confidence score to every answer. Responses take 150 to 250 milliseconds end to end. Google priced input at $0.10 per million tokens with no charge for output, which it says is about five times cheaper than Flash-Lite. The project began as an internal DeepMind prototype, and a small team shipped the model and API about six weeks later. Independent results are more mixed. Artificial Analysis gives Decision 1 a score of 66 on its Decision Index, tied with Jev 1.14 and seven points behind Drex Max v1.1 and GLiDE v1.2. At $345 per million tasks, it costs more than Jev ($156) and Drex Max ($103), and its average of 211 milliseconds per task is slower than Jev's 92 milliseconds. Google's own tests show Decision 1 beating Jev 1.14 on five of seven shared benchmarks, including 92.2% against 90.0% on JevBench. At the AI Engineer event, Cloudflare described a hybrid design in which fast decision models handle tool selection and log extraction while a large reasoning model handles diagnosis. In its demo, that cut six LLM calls to one and roughly 40,000 tokens to 1,800. For Australian businesses and professionals, the immediate impact is limited. Teams running high-volume classification, routing or moderation could benchmark Decision 1 against their current workloads, but the per-task cost and latency figures suggest judging economics per task rather than by the headline per-token price. Any use involving personal information should be reviewed against the Privacy Act 1988 before production, including where data is processed and how Google handles it. For most businesses, nothing changes yet.