Google Launches Gemini Decision 1 for Parallel Thousand-Item Judgments

Google DeepMind has launched Gemini Decision 1, a decision model powering the Gemini Decisions API, designed for closed-ended judgments such as yes/no, multiple-choice, and scale ratings. A single call can process up to 1,000 independent questions across text, images, audio, video, and PDFs, returning probability scores with confidence in 150-250 milliseconds. The API is available in Google AI Studio via Python, JavaScript, and REST, with Google Cloud support coming soon.

According to Artificial Analysis, Gemini Decision 1 scores 66 on the Decision Index v2.1, tying with Elephant 9B and Jev 1.14 but trailing leaders Drex Max v1.1 and GLiDE v1.2 by 7 points. It leads on knowledge and reasoning benchmarks but costs about $345 per million tasks, more than triple Drex Max v1.1, and lags notably on the AA-Paperclip extraction benchmark. Google has internally deployed it for dynamic model routing in Gemini Notebook, design prototyping in Stitch, and search safety review, validating its feasibility for high-concurrency real-time decision scenarios.

As the first decision-model category from a major vendor, Gemini Decision 1 pushes closed-ended batch judgment from rule engines and single-task models toward a multimodal, low-cost, low-latency paradigm, potentially reshaping development workflows for query routing, content grading, and policy compliance. However, cost and accuracy gaps suggest competition in this space is just beginning.