Close-up of a hand reviewing a written plan on a clipboard, evoking a decisive choice.

Google’s Decisions API scores a thousand questions in one pass

Google’s new Gemini Decision 1 model returns probabilities and confidence scores in a single forward pass, DeepMind ships AlphaGenome 2, and AWS releases a free local emulator.

Google today introduced the Gemini Decisions API, powered by a new model, Gemini Decision 1. Instead of generating answer text token by token, it scores candidate options in a single forward pass and returns probabilities plus a 0–1 confidence score. One shared multimodal input — text, images, audio, video or PDFs up to 256K tokens — can be checked against up to 1,000 yes/no, multiple-choice or rubric questions. Google quotes 150–250 ms end-to-end latency and prices the model at $0.10 per million input tokens, with no output tokens billed. That makes it a cheap control plane for routing, policy checks and grading, with a confidence score so ambiguous cases can go to a fallback or a human.

What else matters

AlphaGenome 2 predicts how DNA is read in specific cells

Google DeepMind announced AlphaGenome 2, which predicts how DNA variants affect a chosen cell type or state rather than a tissue average. Trained on more than 100 million cell profiles across 68 tissues, it identified 30% of causal variants on a 28-type immune-cell set, nearly twice the earlier model. It is available for non-commercial research via the AlphaGenome API.

AWS Emulator lets you test cloud APIs on your machine

Amazon released AWS Emulator, a free local tool that accepts the same AWS SDK and CLI calls you already use, with no AWS account or deployment required. Supported operations match AWS results and errors; others return 501. State stays in memory until you stop it, and the source is published under a CC0 public-domain dedication.