Google launches Gemini Decisions API for fast multimodal decision‑making
What happened
Google announced a new Gemini Decisions API built on the Gemini Decision 1 model. The service can process a single multimodal input—text, images, audio, video or PDF—along with up to 1,000 independent questions in one request. It returns probability and confidence scores for each question without generating output tokens, delivering end‑to‑end latency of 150‑250 ms and costing $0.10 per million input tokens.
The API supports three question types: yes/no (noul), multiple choice (choice) with 2‑255 options, and ordered rubric scoring (score). Users can combine modalities, for example pairing a call recording with account data or a scanned claim with damage photos, and control image or video resolution to manage token usage. Responses include a confidence value from 0 to 1, enabling developers to route uncertain cases to human review or fallback logic.
Google cites internal use cases such as real‑time model routing in Gemini Notebook, rapid wireframe generation in Stitch, instant emoji selection in a chat experiment, and safety checks in Search. The Decisions API is available now through Google AI Studio, with a no‑cost tier for prototyping and SDKs for Python, JavaScript and REST. Integration partners include OpenRouter, LangChain, Vercel and others, with broader cloud support planned.
Why it matters
The ability to obtain calibrated, multimodal decisions in milliseconds opens new possibilities for automation in Rwanda and the wider African market. Small businesses that rely on document processing, visual inspection or audio analysis can replace custom classifiers with a single API call, reducing development time and operational costs. For example, a logistics firm could evaluate delivery photos and accompanying notes to flag damaged parcels, while a fintech startup might assess loan documents and voice recordings for compliance in real time.
Because the service charges only for input tokens and does not bill for output, it is financially attractive for startups with limited budgets. The confidence scores also provide a safety net, allowing teams to set thresholds and involve human reviewers only when the model is uncertain. This blend of speed, cost efficiency and transparency can help African enterprises scale AI‑driven workflows without heavy investment in specialized talent.
Key takeaways
- Gemini Decisions API delivers multimodal yes/no, choice and scoring answers in a single fast pass.
- Pricing is based solely on input tokens, with no charge for output, making it cost‑effective for high‑volume use.
- The API’s confidence scores enable automated routing while preserving a human‑in‑the‑loop option for ambiguous cases.
AI-assisted summary published by the Diolichat Team. See our editorial policy. Spotted a mistake? Email info@diolichat.co.rw.