MLX
Named by 36 kept items in this workspace.
21 items name it under frameworks and tools and 15 under classification.technologies.
Search every workspace for MLX. It matches the name as written, so it can find other things that share it.
The items that name it 36
Apple array framework; recent kernels speed batched decoding 2-3x and fused SDPA 1.4x.
Native LFM3 support on Apple hardware.
Community MLX quants of the model are available.
Apple silicon framework underlying mlx-lm-lora (named in the tool name MLX-LM-LoRA).
Apple-silicon framework; a user ran Flint with int4 MLX on an M5 Max.
Backend for running open-weight models locally through AnyLanguageModel (MLXLanguageModel).
Framework used to run the 4-bit TensorFold recipe
Apple runtime for running Muse Spark Open on Mac Studios
Apple framework used to run Parakeet locally in BeamSpeak
Optimized inference for Apple Silicon
Inference backend with day-one support; serves on port 8080
Used to serve the model efficiently
Apple framework used to run DSpark speculative decoding on M5 Max, reaching 3.13x speedup
Apple Silicon inference framework used for DSpark speedup benchmarks
Apple framework used for on-device inference measurements on an M5 Max.
Project, alongside llama.cpp, that has helped make local inference a practical option for everyday use.
Apple's ML framework, mentioned as a possible alternative runtime for Apple's Neural Engine models
Apple Silicon training backend; drives Soup's live dashboard, tracker and soup ui, but ignores several validated config options
Apple's machine-learning framework for its own Apple Silicon chips, enabling local AI inference without cloud costs; unified memory architecture suits mid-sized models (7B-70B) on M series chips
Referenced as an alternative framework; TurboFieldfare is model-specific rather than a wrapper around MLX or llama.cpp
Apple framework; model is optimized to run on two Mac Studios with M5 Ultra
All technologies in this workspace.