Laya on MLX
laya-mlx is an independent MLX port of Laya that runs the original weights on the Apple Silicon GPU with no PyTorch or Transformers runtime. It keeps upstream prompt formatting, calibration and output schema, and it is the base for many Mac demos, agents and benchmarks.
260 projects
laya-plugin
Andrei-cloud
Provides an offline Laya decision plugin for coding agents through MCP and native integrations
laya-explore
Lewiscowles1986
Explores running Laya locally, including an MLX demo for Apple Silicon
laya-opencode-router
Wauuuuuu
A macOS OpenCode plugin routes tasks using local Laya classification and optional MLX models
hermes-laya-arbiter-integration
osviel91
Connects Hermes Agent to Laya/Arbiter decision tools through MCP
jev-clean
chengyixu
A macOS cleanup tool that uses Laya-MLX to investigate System Data and propose files for review
system-one-playground
goodboybeau
A macOS playground comparing Laya, Decider, Kev and Jev on accuracy, calibration and performance
laya-eval
imaddde867
An evaluation compares local Laya with jevmlx and hosted Jev on public cases
laya-mlx-http
mjrussell91
Serves the Laya MLX question-answering model through a local FastAPI endpoint
laya_ai_model_demo
lalithdabilpuram01
Tests Laya on labelled support emails and measures inference latency across local devices
ai-sec-lab
shsingh
A Mac-based Kubernetes lab demonstrates prompt-security defenses using Laya and NOVA
local-jev-bench
tak-bro
Benchmarks local Jev-style decision models, including Laya, on Apple Silicon
laya-mcp
kennedyshead
Serves local Laya routing decisions to coding agents through an MCP server
laya-api
gugaucb
Serves Laya and Laya-MLX models through OpenAI- and Anthropic-compatible APIs
sys1rust
kgarg2468
A Rust and MLX service runs Laya decision models locally and exposes the System 1 API
notchvoice
Starlord-999
A hands-free Mac voice assistant that uses Laya-MLX for on-device decisions
laya-test
psenger
Tests Laya decisions, confidence calibration and latency on Apple Silicon GPUs
laya-stack
spaxfiz
Provides an MLX inference studio, evaluation benchmarks, and data workflows for Laya decision models
laya-service
FnSGit
Serves local Laya inference through a FastAPI endpoint compatible with the Jev protocol
laya-01
D-Ajay-Kumar
Benchmarks and fine-tunes Laya-format decision models for column relevance and yes/no tasks
LAYA_finetune
ameerhmz
Fine-tunes Laya into a multilingual, non-autoregressive decision model and provides deployment formats
laya-service-desk-demo
choas
Demonstrates a local Laya-powered customer service bot using prepared replies and human handoffs
local-systemone
axiomantic
Provides a self-hosted System 1 decision daemon and MCP server with a native Laya backend
local-settings-slm
AkhileshManda
Uses Laya to generate Android settings decisions for distilling into an on-device Qwen model
LocalFirstGmailIntelligence
karan19
Sets up local Gmail intelligence models on Apple Silicon, including Laya-MLX
Install
Requirements from the laya-mlx README: Apple Silicon, Python 3.11+, macOS 14+.
pip install laya-mlx
Pre-converted FP16 checkpoints are on Hugging Face: aac6fef/laya-mlx, aac6fef/laya-multilingual-mlx and aac6fef/laya-typed-decisions-mlx. You can also load the original convaiinnovations/laya IDs.
Router and CLI
The port includes the upstream language router and presets:
from laya_mlx import Router, triage_questions
router = Router(dtype="float16", max_loaded=2)
result = router.predict({"message": "发票被重复扣款，请退款。"}, triage_questions())
print(result["routing"]) # multilingual
For the terminal demo, where Laya plays Snake:
pip install 'laya-mlx[demo]'
hf download aac6fef/laya-multilingual-mlx
laya-snake
Reported numbers
On an M3 Max, FP16, one short question: 13.42 ms P50 for the 421M English model and 7.39 ms for the 322M multilingual model. The README states that all three checkpoints matched the upstream answer on 63/63 validation questions in FP32 and FP16. That measures port fidelity, not task accuracy.
Projects built on laya-mlx
- laya-ultrafast: a port of browser-use/jev-ultrafast that makes browser decisions locally.
- laya-mcp (wsargent): an MCP server with triage, guard, moderation and email tools.
- laya-fast: adds Neural Engine execution next to the MLX GPU.
- laya-jev-lab and jev-laya-benchmark: independent Jev comparisons that use laya-mlx as the Laya side.
- laya-vs-jev: Laya and Jev playing Chrome's T-Rex game side by side.
Caveats
This is not an official Convai Innovations release. Training and fine-tuning stay in the upstream project. BF16 can be requested but is not part of the published validation.
More ways to use Laya