So how do your questions turn into a decision?

Instead of asking a model to describe a photo, you give it a checklist of yes/no, multiple-choice, and rating questions so it knows what to check and what action to take.

Here, Gemini Decision 1 scans 100 delivery photos in 4.5 seconds (nearly 4x faster than Gemini 3.5 Flash-Lite), answering 16 questions per photo at once:

📦 Is the box damaged? (yes/no)
🌧️ Is it under a porch? (yes/no)
⭐ What's the delivery quality? (0 to 3 rating)
⚡ What happens next? (routes to a claim review)