Google launched the Gemini Decisions API on November 11, letting developers score up to 1,000 fixed-answer questions across text, images, audio, video, and PDFs in one request.

The tool uses Gemini Decision 1 and returns probabilities and confidence scores for tasks such as policy checks, model routing, and content grading.

Google said median end-to-end latency was 150 to 250 milliseconds for contexts of up to 16,000 tokens, and performance stayed nearly flat across one to 100 questions.

Pricing is US$0.10 per million input tokens, with no output token charges.

Developers can access the API through the Gemini API in Google AI Studio, which includes a no-cost tier. Support for Gemini in Google Cloud will be available soon.

The use cases overlap with content moderation. OpenAI offers a moderation API guide, while Anthropic covers the topic in Claude documentation.

Model routing is also a topic Oracle describes in its large language model, or LLM, routing documentation.

🔗 Source: Google

Recent Google developments