Laya Mlx

Laya Mlx is a native MLX runtime for Laya typed decision models running locally on Apple Silicon.

Laya Mlx is an open-source, native MLX runtime designed to run Laya typed decision models locally on Apple Silicon. It provides local inference for constrained questions like choices, rubric scores, and probabilities in a bidirectional forward pass without token-by-token decoding. The software is built for developers working with Apple Silicon hardware who need fast, local decision-making capabilities. It supports pre-converted FP16 checkpoints from Hugging Face and includes a Python API as well as command-line tools. Key components include support for English and multilingual checkpoints, temperature calibration, prefix caching, and optional compilation for optimized throughput.

Laya Mlx at a glance

Best for: Software engineer
Runs on: macOS, CLI, self-hosted, open source
Pricing: Free
Released: September 2026
Category: AI agent tool
Website: github.com/mizorewww/laya-mlx

Key features

Run Laya typed decisions locally on Apple Silicon
Support choice, score, and probability outputs
Load pre-converted FP16 checkpoints from Hugging Face
Provide a Python API for predictions
Include command-line demo tools
Support prefix caching and prompt compilation
Handle language routing and preset workflows

Use it for

Running local decision models in Apple Silicon environments
Evaluating choices, rubric scores, and probabilities
Processing batch prediction queries via Python API
Shortlisting large choice sets using cosine similarity

Alternatives to Laya Mlx

MLX Direct Decider — Runs local MLX decisions without fine-tuning or extra weights.
Rapid-MLX — Run 250 MLX-native models locally on Apple Silicon with high inference speeds.
MTPLX — It runs local models twice as fast via native MTP speculative decoding.
MLX LM — Runs LLMs and fine-tuning natively on Apple silicon hardware.
Strata Mlx — Runs 125B MoE models on Apple Silicon by streaming experts directly from disk.
OpenSquilla — Maximizes AI intelligence density while minimizing token costs through smart routing.
Union.ai — Runs machine learning tasks natively on Apple Silicon GPUs using Metal.
MLflow — Complete observability and evaluation for production AI apps.
Melra — Governs AI agent tool calls with local policies, approval gates, and verified receipts.
Phocinae-Largha-150M-v1（斑海豹 Largha） — Runs locally with zero generated tokens and high decision speed.
Agent Squad — Runs multi-agent orchestration in Python, TypeScript, and locally on Apple devices.
MyAgent — It features built-in permission controls, long-term memory, and skill self-evolution.

All Laya Mlx alternatives