328 points69 comments
LLMs now run on old TV boxes and decade-old iPads
What is I'm making LLMs work on small and old devices?
smolmaxxing finds the biggest language model a low-end device can actually run and then pushes it further with custom kernels, smarter memory handling and weight streaming. It demonstrates a 0.8 B-parameter model at 16 tok/s on a cheap TV box and a 1.7 B model at 25 tok/s on a 2013 iPad.
Who it is for
If you have spare hardware, you can get useful LLM inference without buying a new GPU.
The story so far
Early Signal first spotted I'm making LLMs work on small and old devices on November 12, 2026, with 328 points. On its busiest day it gained 83 points. Early Signal files it under LLMs.