laya
The most-liked open-weight decision model of the post-Jev wave, a 421M ModernBERT-large encoder. The base release of a three-model family.
Community card
What it is
Laya is a 421M-parameter encoder on ModernBERT-large, and the most-liked model in this
directory by a wide margin; the current count is in the panel alongside. It anchors a family: laya-typed-decisions is the
variant tuned on the community benchmark set, and laya-multilingual swaps the base for mmBERT.
Why it draws the attention
Those likes are a measure of demand: people want an open answer to Jev that runs on their own hardware, and an encoder this size runs on a CPU. It is also the root of the widest family in the directory, with GGUF, MLX, ONNX, Core ML and LiteRT builds already published by other accounts.
Treat the base laya checkpoint as the general-purpose starting point and
laya-typed-decisions as the one to compare against published
figures, since that is the variant whose card carries them.
A public number is a shortlist, not a decision
The only benchmark that settles which model you ship is one built from your own historical decisions. We freeze that test set, run the candidates against it, and hand back the threshold your error budget supports.