← November 11, 2026 briefing · added in the 04:00 KST update

Google launches Gemini Decisions API, built on its new Gemini Decision model, for fast, calibrated multimodal judgments

Announced November 11, 2026

Verified (fact)

Google introduced the 'Gemini Decisions API' on its official blog. The API runs on 'Gemini Decision', a new Google model. It takes text as well as other inputs such as images, makes judgments quickly, and returns calibrated confidence scores with them. The collected excerpt does not include details such as pricing, supported input formats, or benchmark figures.

Needs fact-check (claim)

Google described Gemini Decision as its state-of-the-art model.

Why it matters (interpretation)

Nace AI's 'Drex', Amazon's 'Strands Decider', and Kakao's 'Kanana-2-v-decide' came out recently, and LangChain has now made decision models a standard feature. Models that skip long answers and quickly return just a choice and its probability are spreading fast for agent classification, routing, and evaluation. With a frontier company like Google releasing a dedicated API, this category is more likely to become its own product line. Replacing existing LLM calls with these models can cut cost and latency.

Sources

More about Google