community Reddit LocalLLaMA 2026-09-21T12:25 正文 thin 0 words

[Splash Engine] Qwen3.8-27B in native 8-bit at 37–55 tok/s on Apple Silicon: Extending Splash to Q8, 256k context scaling, and the "Reasoning Cliff"

英文标题：[Splash Engine] Qwen3.8-27B in native 8-bit at 37–55 tok/s on Apple Silicon: Extending Splash to Q8, 256k context scaling, and the "Reasoning Cliff"

https://preview.redd.it/nulsv53o8vqh1.png?width=4500&format=png&auto=webp&s=74765dbd409f4c221640f9f6000a685f6fdbb242 Spent weekend benchmarking the Splash engine (by Incoai) and extending its architecture to native 8-bit on Apple Silicon (M5 Pro, 64 GB unified memory). Splash is a compiled C++

打开原文相关 2 · 机会 0

中文

https://preview.redd.it/nulsv53o8vqh1.png?width=4500&format=png&auto=webp&s=74765dbd409f4c221640f9f6000a685f6fdbb242 Spent weekend benchmarking the Splash engine (by Incoai) and extending its architecture to native 8-bit on Apple Silicon (M5 Pro, 64 GB unified memory). Splash is a compiled C++

中文全文翻译正在补齐，可先查看下方英文原文。

[Splash Engine] Qwen3.8-27B in native 8-bit at 37–55 tok/s on Apple Silicon: Extending Splash to Q8, 256k context scaling, and the "Reasoning Cliff"

中文

English / 原文