Google announced the Gemini Decisions API, a new endpoint powered by the Gemini Decision 1 model (gemini-decision-1). Instead of generating text token by token, the model evaluates up to 1,000 closed-set questions in a single pass, returning probabilities and confidence scores.
How it works
The API accepts a shared multimodal input (text, images, audio, video, or PDFs, up to 256K tokens) and up to 1,000 independent questions of the yes/no (noul), multiple choice, or rubric scoring type. End-to-end latency is 150-250 ms, and pricing is $0.10 per million input tokens, with zero output tokens billed.
According to Google, the model leads in agent routing, reasoning, and multimodal understanding benchmarks, with the lowest calibration error. It is nearly 5 times cheaper than Gemini 3.5 Flash-Lite on shared test suites.
Why it matters
This is useful for AI agents that need to make many fast decisions: tool selection, policy checks, model routing, and context grading. Google teams already use it in products like Gemini Notebook and Stitch to reduce latency.
Available via the Gemini API in Google AI Studio, with a free tier for prototyping. Integrations with OpenRouter, LangChain, and others are underway.
Official source: Google blog.
By Grok
