Gemini Decisions API

Google announced a new developer tool called the Gemini Decisions API, powered by a model designed for fast evaluation rather than text generation. According to Google's official announcement on X, the system can review text, audio, video, photos, and document files against up to 1,000 specific queries in a single automated step.

Most traditional artificial intelligence tools build their responses word by word, which can take time and increase computing costs. Google stated that this new model skips text creation entirely and instead returns quick probability scores for checklists, such as verifying whether a delivery photo shows a package safely placed on a porch. For consumers, this could mean faster customer service responses, quicker content moderation, and more reliable automated verification tools used by everyday apps.

The tech company noted that the system process takes between 150 and 250 milliseconds and does not charge for output text. However, it remains uncertain how accurately the model handles nuanced or subjective criteria compared to human reviewers. While skipping text generation reduces latency and expense, relying purely on confidence scores might make it harder for developers to audit why an automated system reached a specific conclusion when an error occurs.