Google launches Gemini Decisions API, scoring up to 1,000 yes/no, choice or rubric questions in one pass
The model returns probabilities and confidence scores instead of generated text, with 150–250 ms latency at $0.10 per million input tokens and no output tokens billed.
Why it matters
Agent builders often call a full generative model just to route a request, check a policy or grade an answer, which adds cost and latency. Gemini Decision 1 instead scores the candidate options in a single forward pass over text, images, audio, video or PDFs of up to 256K tokens. Google says it leads its benchmark suite with the lowest calibration error and is nearly five times cheaper there than Gemini 3.5 Flash-Lite. The model was spotted in API listings earlier today before the official launch. Watch independent tests of its calibration and how developers use it for routing.
Key facts
- Model: gemini-decision-1, in Google AI Studio via the Gemini API
- Up to 1,000 questions per request on inputs up to 256K tokens
- Question types: yes/no, multiple choice (2–255 options), ordered rubric
- 150–250 ms end-to-end latency
- $0.10 per 1M input tokens; zero output tokens billed
- Nearly 5x cheaper than Gemini 3.5 Flash-Lite on Google's benchmark suite
How the story developed
- 11 Nov, 19:20 UTC Breaking Google launches Gemini Decisions API, scoring up to 1,000 yes/no, choice or rubric questions in one pass
- 11 Nov, 20:21 UTC Confirmed Google launches Gemini Decisions API, scoring up to 1,000 yes/no, choice or rubric questions in one pass