GoKawiilGoogle launches Gemini Decision 1 API for fast multimodal scoring tasks
Google has introduced Gemini Decision 1 via a new Decisions API, which scores inputs directly without generating token-by-token text output. It handles text, images, audio, video and PDFs in a single pass, supports yes/no, multiple-choice and rubric-style questions with confidence scores, and keeps latency around 150-250 ms end-to-end regardless of input volume up to 100 questions per request.
GoKawiil's interpretation of the reporting above, not reported fact.
By skipping output-token billing and decoding, Google positions this model as a cheaper, faster alternative for high-volume classification and scoring tasks than general-purpose chat models like Gemini 3.5 Flash-Lite, according to the company's own benchmark comparison showing nearly 5x cost savings. The built-in confidence scores could let developers automate routing of uncertain cases to human review, which may appeal to industries like insurance or customer service handling large volumes of structured decisions.
- Gemini Decision 1 scores options without generating output tokens, cutting latency to 150-250 ms end-to-end
- It processes multiple modalities (text, images, audio, video, PDFs) together in one forward pass
- Google claims it is nearly 5x cheaper than Gemini 3.5 Flash-Lite on a shared benchmark suite
Source: blog.google — Min Choi, 2026-11-11
Published there as: “Gemini Decisions API”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.