Laya-CoreML: 4.98 ms decisions on the Neural Engine, 2.78x better energy than MLX

# Laya-CoreML

**Open-weight typed decisions on Apple Silicon. Core ML, Neural Engine, zero generated tokens.**

[PyPI](https://pypi.org/project/laya-coreml/) · [Hugging Face weights](https://huggingface.co/aac6fef/laya-multilingual-coreml-ane) · [中文](https://github.com/mizorewww/laya-coreml/blob/main/README.zh-CN.md)

A real Laya model plays Snake locally, with visible probabilities, score, length,
latency and safety interventions. The GIF replays a recorded Core ML run at **1× speed**.
The game uses explicit planner features and a visible cycle safety layer.

The complete active Snake loop sustained **49.1–50.0 decisions/s** across three
uncapped 600-step episodes, with zero deaths and two safety interventions.
[Game-loop timings and paced-rate limits](https://github.com/mizorewww/laya-coreml/blob/main/docs/SNAKE_BENCHMARKS.md)
include rendering serialization; terminal painting is excluded.

**One short multilingual decision: 4.98 ms P50 / 5.31 ms P95 on M3 Max with ANE FP16.**
The same experiment measured **2.78× better whole-system energy per decision** than
compiled MLX FP16. A separately validated W8 palette variant reached 4.88 ms and
3.19× energy improvement. These are single-question results, not full Snake frame
times; the requested 10× improvement was not achieved.

## Run the demo

Apple Silicon · macOS 15+ · Python 3.11–3.13.

```bash
pip install 'laya-coreml[demo]'
hf download aac6fef/laya-multilingual-coreml-ane --local-dir models/snake
laya-coreml-snake --model ./models/snake
```

Download once, then play offline. No PyTorch, Transformers or MLX is needed for
inference. The terminal needs 104 columns × 35 rows. Space pauses; ↑/↓ changes
speed; R resets; Q quits. First-time Core ML initialization can take tens of seconds.

[Controls, recording and video e