Google launches Gemini Decisions API for bounded choices and model routing

Read the original article →

Google announced the Gemini Decisions API on November 11, 2026, making its Decision 1 model available through the Gemini API in Google AI Studio. The endpoint evaluates predefined choices rather than writing a free-form answer.

What Google announced

The API accepts text, images, audio, video and PDFs, with up to 1,000 independent questions against a shared input. Questions can use yes/no checks, multiple-choice options or an ordered scoring rubric. Responses include probabilities and confidence scores.

Google lists pricing at $0.10 per million input tokens, with no output-token charges and a no-cost tier for prototyping. The shared input is billed once per request; question instructions and criteria still add tokens.

Google reports 150–250 ms end-to-end latency in its described tests with contexts up to 16K tokens. That is a vendor measurement, not a guarantee for every workload. MintedBrain has not independently tested the API.

MintedBrain analysis: start with a reversible decision

A fictional support workflow could classify a message as “billing,” “technical” or “needs review,” then send it to the appropriate queue. This illustrates a possible use, not a demonstrated result.

Before deployment, define what each option means and evaluate representative examples, including ambiguous requests. Set a fallback for uncertain cases and measure wrong-route rates. A confidence score is a model estimate, not proof that the choice is correct. Keep consequential approvals behind appropriate checks.

Explore MintedBrain tutorials for workflow-building lessons and the task directory to compare tools by the job you need done.

Source: Google’s launch announcement, November 11, 2026.

References

This article was originally published at Google. For the full piece, read the original article.

Discussion

  • Loading…

← Back to News