Google launches Gemini Decisions API for classification and scoring

Google has launched the Gemini Decisions API, a specialized endpoint for classifying and scoring data, capable of processing up to 1,000 questions per request. Pricing is set at $0.10 per million input tokens with no output charges. The API uses the dedicated Gemini Decision 1 model to handle text, images, audio, video, and PDFs, supporting binary judgments, multiple-choice selection, and rubric-based scoring. Response times range from 150 to 250 milliseconds, with input limits of 256K tokens. This release follows similar offerings from TypeSafe AI, OpenAI, and Microsoft. Google claims its model outperformed OpenAI's GPT-6 Luna on all seven shared benchmarks, including a 92.2% accuracy rate on JevBench, while also demonstrating the lowest prediction calibration error among competitors.