• Sources: primary, discussion, discussion
  • Summary: Google released the Gemini Decisions API and the gemini-decision-1 model on 2026-11-11. One request takes one multimodal input of up to 256K tokens, covering text, images, audio, video or PDF, plus up to 1,000 questions of three types: noul (yes/no), choice (2 to 255 options) or score (an ordered rubric). The model scores all options in one forward pass and returns a probability and a confidence for each. Google cites 150 to 250 ms end-to-end latency. Pricing is $0.10 per million input tokens with no output charges, and Google says Gemini Notebook uses the API for model routing and Search uses it as a safety layer for MCP tools.
  • Why it matters: Routing, tool filtering and policy checks in agent pipelines can use a fast scoring call with confidence scores instead of a generative call.
  • Comments: Simon Willison says the API is almost identical to Jev's, down to reusing Jev's term noul for yes/no questions. The Hacker News thread drew 202 points and 68 comments.