AI/ML●●●Banger
SwiftLM – Qwen Chat on iPhone, 100B+ Moe on M5 Pro 64GB (Native Swift)
Native Swift inference with SSD streaming runs 100B MoE models without kernel panics.
WizardryNiche Gem
aegis_camera
123mo ago
Back to browse
GitHub Repository
MLX-compatible REAP for pruning MoE models on Apple Silicon
8 starsPython
Ported Cerebras REAP to MLX – Prune MoE Experts on a MacBook
by egesabanci·Jun 2, 2026·2 points·0 comments
AI Analysis
●●SolidNiche GemBig Brain
MoE pruning on MacBook without CUDA or PyTorch dependency stack.
Strengths
- •MLX-only runtime avoids importing Torch, vLLM, or plotting libraries.
- •Adapter architecture isolates model-family differences cleanly.
- •Structured telemetry writes validation metrics and pruning decisions per run.
Weaknesses
- •Port of existing Cerebras research, not novel algorithm or architecture.
- •Narrow audience limits broader appeal beyond MoE researchers.
Category
Target Audience
ML researchers and engineers working with MoE models on Mac
Similar To
Cerebras REAP · MLX-LM
Similar Projects
AI/ML●Mid
Phase Router – capacity-aware routing for MoE
Specialized routing logic for MoE models without a demo or benchmarks.
Niche Gem
TSltd
512mo ago
AI/ML●●●●Gem
Run a 120B-parameter MoE on Android mid-range phone CPU-only llama.cpp
Runs 60GB models on 12GB phones by streaming experts from flash, not RAM.
WizardryZero to OneBig Brain
Helldez
1016h ago
AI/ML●●●●Gem
ROLV – 20x faster MoE FFN inference on Llama 4 Maverick vs. cuBLAS
20x faster MoE inference on existing hardware with hash-verified output correctness.
WizardryZero to One
heggenhougen
114mo ago
AI/ML●●●Banger
Getting GLM 5.2 running on my slow computer
Streams 744B MoE experts from disk to run on 25GB RAM—no GPU, pure C.
WizardryBig BrainZero to One
vforno
93624011d ago
AI/ML●●Solid
mlx-chronos - benchmark MLX inference engines on Apple Silicon
Standardized MLX benchmarking when everyone's currently comparing engines manually.
Niche GemBig Brain
igurss
2024d ago