Thursday, 12 November 2026, story 1 of 7
Google launches Decisions API to score up to 1,000 questions per call
Gemini Decision 1 returns confidence scores in 150-250 ms for $0.10 per 1M input tokens, with no output-token charge, Google says.
What happened
Google DeepMind has released the Gemini Decisions API, built on Gemini Decision 1. It is available now through the Gemini API in Google AI Studio. One request takes a multimodal input of up to 256K tokens and up to 1,000 yes/no, multiple-choice or rubric questions, and each answer carries a confidence score from 0 to 1.
Google lists $0.10 per 1M input tokens, no output-token charge and a no-cost tier. It gives 150-250 ms end-to-end latency and says the setup is nearly 5x cheaper than Gemini 3.5 Flash-Lite with the Interactions API on a shared benchmark suite. Integrations are rolling out; Google Cloud support is promised.
What it means
The target is narrow: swap a generated answer for a scored one wherever the answer comes from a fixed set, such as routing, policy checks or grading. The confidence score is the practical part, since code can send ambiguous cases to a fallback or a human reviewer, as Google suggests. Billing is also easier to predict: the shared input is billed once per request, and each extra question adds only its own instruction and criteria tokens.
Read the numbers with care. The 5x figure is Google's, from its own benchmark suite. The flat-latency claim covers contexts up to 16K tokens and 1 to 100 questions, not the 1,000 maximum. Test accuracy and calibration on your own data before replacing an existing classifier.
What to do: If you route queries or run policy checks with a generative call, test Gemini Decision 1 on the no-cost tier against your own labelled cases, checking calibration.
Sources
Filed under Models. Published .
"What it means" is analysis, not reporting. Found an error? Write to hello@vibitely.com.