Google Gemini Decision 1 released: does not generate verbatim, judges 1000 questions in one request
Google released its first decision model, Gemini Decision 1: it does not generate word by word, and can score up to 1000 questions with a single forward pass, with an end-to-end latency of 150 to 250 milliseconds and only $0.10 per million input tokens; it achieved first place in 7 out of 9 benchmarks, with a highest accuracy rate of 92.2%.