AINews wire, 11 Nov 2026
Google launches Gemini Decisions API for yes/no, choice and rubric checks
Reported by Google – The Keyword, Simon Willison

Google has launched the Gemini Decisions API, a developer interface for closed-set decisions. It runs on a new model called Gemini Decision 1 and is available through the Gemini API in Google AI Studio. Instead of writing answer text one token at a time, the model scores fixed options in a single pass and returns probabilities and confidence scores.
Developers can send one multimodal input made of text, images, audio, video and PDFs, up to 256K tokens, and attach up to 1,000 independent questions to it. Three question types are supported: yes/no checks, multiple choice with between 2 and 255 options, and rubric scoring on an ordered scale that starts at 0. Each answer carries a confidence score from 0 to 1, so code can send unclear cases to a fallback or a person.
According to Google, median model compute time is 50 to 100 ms, or 150 to 250 ms end to end, and latency stays nearly flat from 1 to 100 questions per request. Google also says Gemini Decision 1 leads its benchmark suite with the lowest calibration error, and that it is nearly five times cheaper than Gemini 3.5 Flash-Lite with the Interactions API across that suite. These are Google's own claims and have not been checked independently.
The price is $0.10 per million input tokens, with no output tokens billed. The shared input is billed once per request, and each question adds only its own instruction and criteria tokens. Google says there are no separate charges for thinking tokens, audio transcription or OCR. A no-cost tier is available in Google AI Studio for prototyping.
Google names several internal uses. Gemini Notebook uses the API as a router that sends queries to the right model tier, which Google says can cut latency by up to 40 percent for everyday queries. Stitch uses it for a rapid design canvas, CC uses it to choose emoji reactions from more than 200 options, and Google Search uses it as a safety layer for MCP tools. Support is also coming to several developer tools, including OpenRouter, LangChain, Vercel, LiteLLM, Pydantic AI, Pipecat, LiveKit and OpenCode, and to Gemini in Google Cloud later.
Simon Willison, in a link post, describes the API as almost identical to a rival decision API from another provider, which he calls Jev, and notes that it reuses that provider's term for yes/no questions. He says the price matches the GPT-6 Luna decision model from OpenAI, while Jev has the lowest rate of the three, roughly 4.2 cents for every million input tokens. In his view, Jev and OpenAI's model handle text, while Gemini Decision 1 also covers audio and video. He tested it on a video file and found that Gemini gave less than full confidence on yes/no answers, where the other two returned 1.0 or 0.0.
Several points are still open. The sources do not give a general release date beyond the no-cost tier, a timeline for Gemini in Google Cloud, or independent test results for the calibration claims outside Google's own benchmark suite.
Enjoying this? Get the AI brief
Every Thursday: AI that matters for your home.
✓ Free · seven editions · one click to leave
How each source covered it
Original reporting by the sources above. Summary by smart-home.space.