Google now has its own Jev

🤖 Google now has its own Jev

Google has introduced the Gemini Decisions API—an endpoint powered by the gemini-decision-1 model for rapid "closed" decisions: yes/no, selection from 2–255 options, and rubric-based scoring. It supports multimodal input: text, images, audio, video, and PDFs up to 256K tokens. It handles up to 1,000 questions per pass, providing probabilities and confidence scores without token generation. Pricing: $0.10 per 1 million input tokens, $0 for output; latency is 150–250 ms.

🌍 For agents, this acts as a "control layer": instead of using a generative LLM for every classification, routing, or policy check, it offers a cheap endpoint that evaluates candidates in a single forward-pass. Inside Google Notebook, it reduces latency by up to 40%.

👤 Routine "yes/no" checks in agents are reportedly about 5 times cheaper than standard LLMs, according to Google. Independent calibration benchmarks are not yet available—test on your own data.

Source 1: https://blog.google/innovation-and-ai/technology/developers-tools/gemini-decisions-api/