AI
Splash Engine Accelerates Local LLMs on Mac
Inco AI has released Splash, an open-source inference engine for Apple silicon that uses speculative decoding and Metal kernels to accelerate local large language models.
October 1, 2026, 11:39 AM·Хабр

Inco AI has released Splash, an open-source inference engine for Apple silicon that uses speculative decoding and Metal kernels to accelerate local large language models.

Fedora Linux Asahi Remix 45 beta introduces upstream support for Apple M3 processors and adds hardware-accelerated video decoding.
Experimental tests of NVIDIA DLSS 5 on Apple M5 Pro chips show improved image quality but a drop in frame rate to 2.32 FPS and ten-fold increase in input lag.

Developers have unofficially ported Nvidia's DLSS 5 to Apple Silicon, but the M5 Pro chip struggles with high input lag and low frame rates.

The Asahi Linux distribution now supports Apple M3, M3 Pro, and M3 Max chips, enabling Linux on newer Mac hardware with some functional limitations.
