# Splash 1.7.0 Released - Even faster decode on Apple Silicon — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/b2e993b95b168a7b

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-11-18 11:35 ingested 2026-11-19 02:39

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1xzd1lx/splash_170_released_even_faster_decode_on_apple/)

## retained text



```
New version of Splash just released. The biggest improvement from previous version is a noticeable speed-up in decoding. It's even faster now.

BTW, this isn't just another slop engine. It's coming from the researchers behind DFlash and DFlash2, who also operate one of the fastest API providers for open models. These guys clearly know what they're doing.
```



document fetched 2026-11-19 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1xzd1lx/splash_170_released_even_faster_decode_on_apple
