Back to home
mlx-dspark
2 articles tagged with this topic
QwenApple Silicon
Apple Silicon Runs Qwen 27B 3x Faster — Local LLMs Enter Usable Territory
mlx-dspark ports DeepSeek's speculative decoding to Apple Silicon, giving Qwen 27B a 3x speedup on M-series Macs with no quality loss. Local LLMs near
Aug 152 min read
MetaMuse Glimmer
Meta's Muse Glimmer 30B Hits 3.3x Speedup on Mac as Open Source Closes the Gap
Developer A-Rahim uses speculative decoding to push Meta's Muse Glimmer 30B up to 3.3x faster on M4 Pro, with output exactly matching the original.
Aug 122 min read