Google introduced Gemini Decisions API for classification and evaluation. Up to 1,000 questions share multimodal context of up to 256,000 tokens, returning probabilities for yes/no answers, two to 255 choices or ordered rating scales. Pricing is $0.10 per million input tokens with no output charge. The advertised 150–250 ms latency applies to contexts of at most 16,000 tokens and one to 100 questions, not necessarily the maximum workload.
Back to Tech
Gemini Decisions API returns structured judgments instead of open-ended prose
Shared context supports many questions, but the quoted latency applies only to smaller inputs.