Laya (English): NoulXP package (unofficial)
Unofficial. This is a NoulXP package of Laya (English) by Convai Innovations, made by System One Models and published here, in Convai Innovations' organization, on its behalf. Convai Innovations did not make or review it. The original, its card and its own code: convaiinnovations/laya, revision 7b928d828b7b. The original on System One Models: convai-innovations/laya.
If you made this model: this repository is in your organization here; claim it at https://systemonemodels.ai/convai-innovations to run it yourself, or write to [email protected] and we will correct this package or take it down, the same day.
Laya (English) is the English checkpoint at the root of Convai's Laya repository: ModernBERT-large with a decision head that scores one marker per option, trained with RLCD. A NoulXP package runs it on any NoulXP runtime (noulxp run, noulxp serve, the System One Engine) with no model-specific code: nothing in it executes.
Lineage
answerdotai/ModernBERT-large (Answer.AI, Apache-2.0) → Laya (English) (Convai Innovations, Apache-2.0) → this package (the GPU package).
What we changed
- Format: ONNX graph, opset 18, with the NoulXP files that describe the input, calibration and answers.
- Precision: float16 matrix products (as PyTorch's autocast): the encoder's weight matrices stored float16, activations cast at each product; embeddings, normalisation, softmax, the residual stream, the decision head and the rotary angles stay float32.
- Nothing else. The input construction, calibration and answers are the model's own.
noulxp/conformance.jsonlholds what the model's own code answers to NoulXP's 52 requests (recorded on its published weights in float32, on a CPU), and this package reproduces it.
Our measurements (2026-10-05)
Measured by System One Models, not by the maker. Fidelity is measured against the model's own answers: NoulXP's 52-request conformance set and the typed-decisions test split (LocalLLaMA/typed-decisions, 400 requests, 2,000 questions).