AIHacker Newsabout 2 hours ago
Open-source engine runs Gemma 4 in 2 GB RAM
2 min read
TurboFieldfare is a new open-source inference engine written in Swift and Metal that runs a 4-bit Gemma 4 26B-A4B-IT model on any M-series Mac using roughly 2 GB of RAM. The model's quantized weights occupy about 14 GB.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
