Google launches Gemini Decisions API for calibrated choices

Google DeepMind on November 11 released the Gemini Decisions API, powered by a dedicated model called gemini-decision-1 that scores up to 1,000 yes/no, multiple choice or rubric questions in a single forward pass instead of generating answer text token by token.

The target is agent plumbing: tool selection, policy checks, model routing, context grading. In the launch post on blog.google, product manager Min Choi and principal scientist Felix Yu pitch "150-250 ms end-to-end latency" with a confidence score between 0 and 1 attached to every answer. Pricing is listed as "$0.10 per 1M input tokens with zero output tokens billed."

Google's own numbers claim "Gemini Decision 1 leads across the suite, with the lowest calibration error" and is "Nearly 5x cheaper across the shared benchmark suite than Gemini 3.5 Flash-Lite with the Interactions API." Neither claim has been reproduced outside Google yet.

Launch partners include LangChain, Vercel, OpenRouter, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode; Google Cloud support is listed as coming soon. Internally, Google says its Gemini Notebook product uses the model for real-time tier routing and sees "up to 40% lower latency" as a result. The link was moving through a couple of the researchers we track on launch day.