netrunner(@plotarmordev):On a DGX Spark, what you run still decides how fast it goes. For example, just turning on MTP in llama.cpp you can make Qwen3.8 27B on a single Spark far faster. Lots more cases like this everyday. We'll push this box to the limits until a new version is out...

netrunner

@plotarmordev

Cofounder & engineer. Running local AI, building things with it, and sharing what I learn.

가입 April 2022

675 팔로잉 중 7.1K 팬

netrunner@plotarmordev

2026.09.26 12:02

On a DGX Spark, what you run still decides how fast it goes. For example, just turning on MTP in llama.cpp you can make Qwen3.8 27B on a single Spark far faster. Lots more cases like this everyday. We'll push this box to the limits until a new version is out...