> Canonical page: https://posttrainllm.com/inspiration/splash

A note of thanks · Mac runtimes

# Splash

Splash shows what a model-specific Mac engine can gain by specializing kernels, weights, cache, and startup memory. Its published numbers are the team's report, not our own benchmark.

01 · The idea

## What stayed with us

Shape-specific Metal kernels, a model-specific speculative draft, packed weights, cache reuse, and a startup memory plan trade generic model support for Mac-local serving performance.

02 · The local translation

## What we did with it

The retained exercise compares a specialized engine with a general engine on cold and cached latency, prefill, decode, concurrency, memory, output validity, and task completion.

03 · The boundary

## Where the comparison stops

A model-specific runtime can be fast without being a general serving replacement, and tuning cost must be included beside runtime gains.

Source trail

## Follow the work

- Original project [Splash source ↗](https://github.com/incoai/splash)
- Further reading [Launch article ↗](https://inco.ai/blog/splash/)
- Our evidence · study only [Splash model-specific Mac inference ↗](https://posttrainllm.com/studies/splash-mac-inference)

Independent appreciation. The named projects have not endorsed or affiliated with PostTrainLLM.