Laya

Laya is an open System One decision model from ConvAI Innovations, built on a ModernBERT-large encoder rather than a generative LLM. You send a state (text, a JSON object or an array of text) with named questions, and it returns typed answers in one forward pass: the probability that a statement is true, a choice among options you define, or a position on your own rubric, each with a full probability distribution. It writes no text, so answers cannot fall outside the options given. Laya speaks the same request and response format as TypeSafe's System One API, so code written for Jev can switch models by changing the model name. Its authors train it with reinforcement learning against proper scoring rules so its probabilities are calibrated. They also state that the shipped checkpoints are a base for fine-tuning: on their own typed-decisions benchmark the base English checkpoint scores 0.362 accuracy against 0.766 for a fine-tuned one, so measure it on your own decisions before relying on it zero-shot. Use it for short, atomic decisions composed in code, such as ticket routing, intent detection and guardrail checks. Input beyond 512 tokens is cut off, so it suits messages and tickets rather than long documents. Laya is published by ConvAI Innovations at https://huggingface.co/convaiinnovations/laya under Apache-2.0.