Google DeepMind has introduced the Gemini Decisions API, powered by the Gemini Decision 1 model. Designed specifically for closed-ended tasks like multi-choice selection, binary logic, and rating scales, the model can process up to 1,000 independent questions per request across multimodal inputs including text, audio, video, and PDFs.

Rather than relying on autoregressive token generation, Gemini Decision 1 evaluates options directly, returning probability scores in 150 to 250 milliseconds. Google priced the API at $0.10 per million input tokens with no fees for output tokens, stating it operates at approximately one-fifth the cost of Gemini 3.5 Flash-Lite.

The release follows similar decision-focused model launches from startups and major tech providers like OpenAI and Microsoft. The API is accessible via Google AI Studio, with integrations for developer frameworks such as Vercel and LangChain currently rolling out.

Why it matters

  • Offers a ultra-low-latency, zero-output-cost alternative for standard classification and decision logic.

  • Facilitates cheap multimodal input processing across text, audio, video, and document formats.

  • Reflects an industry trend toward specialized, non-autoregressive decision models for software pipelines.

Source: the-decoder.com