Laya: open-source 322M model matches TypeSafe Jev with 21ms local decisions
markjeffrey · x · 2026-09-21
Laya is an Apache-2.0 open-source alternative to TypeSafe Jev: a 322M-parameter, 650MB model that decides instead of chatting — typed answers with real probabilities in 21ms on an Apple M1 Max, roughly 200x faster than a hosted LLM, with zero tokens generated. On the same 500 examples Jev is more accurate out of the box, but Laya matches it on simple questions, runs offline, is free per call, and is fine-tunable. Its author published the underlying work in March 2025, 18 months before Jev launched. Unlike a BERT classifier, the question is set at request time — one System 1 model handles any typed question with no per-task training.
Related event: Open-source Laya emerges hours after Jev goes closed-source(4 posts)→
More from Models
- OpenAI's five biggest failures laid out: pretraining decline, alignment fears, enterprise lost to Claude — teortaxesTex · 2026-11-12
- Fable 5.2 Hits 41.3% on Remote Labor Index, Up From 2.5% a Year Ago — alliekmiller · 2026-11-12
- Beaker automates MIT's weight-space programming: one-neuron edit lifts MATH-500 from 7% to 37% for under $4 — VictorSanh · 2026-11-12
- Researcher says Anthropic dodged questions on Mythos false positive rate; 37,966 findings, 2 exploited — OwariDa · 2026-11-12
- Grok spotted building meeting "Listen to Conversation" feature with live transcript — nima_owji · 2026-11-12
- MoonMath ships uncensored 763B DeepSeek-V4.1 via closed-form weight editing, claims zero capability loss — udmrzn · 2026-11-12