Running Kimi K3 on a M1 Mac
SMRTR summary
Deltafin is a research project that runs Kimi K3, a 2.8-trillion-parameter AI model, on a standard Apple Silicon Mac with just 64 GB of RAM. Because the model only activates a small fraction of its parameters per token, Deltafin reads only the needed experts from local disk or streams them from the internet. Speed is slow — about 16 seconds per token on an M1 Max — but Macs with more RAM or newer chips run it faster automatically.
SMRTR provides this summary for quick context. The original article belongs to Hacker News.
Read the original article