mlx-dspark Update Boosts Apple Silicon Inference Speed
GlennCameronjr · x · 2026-07-18
The mlx-dspark update significantly accelerates model inference on Apple Silicon. Gemma-4 12B now runs 2.1x faster (up from 1.8x) with zero loss in output quality, thanks to new kernels in mlx 0.32.
Additionally, v0.4 introduces the first effectively working speculative decoding on Mac for the ternary model Bonsai-27B, achieving a 93% acceptance rate on coding tasks.
More from Infra
- Encrypted GPT-2 speedrun challenges FHE inference: best is 3.031 s/token, goal sub-1s — tensorqt · 2026-10-27
- Open-source Latch compiles PyTorch into training megakernels, 2.75x faster on H100 — sguada · 2026-10-27
- Stanford launches GPU-centric supercomputer Marlowe for data science and AI research — StanfordHAI · 2026-10-27
- NVIDIA Claims HGX Rubin NVL8 Delivers 10x More Tokens per Megawatt Than B200 — thursdai_pod · 2026-10-27
- Together lets multiple LoRA post-training experiments share the same GPU capacity — togethercompute · 2026-10-27
- Commoditization dilemma: either human labor or GPUs become the commodity — _AustinCalvert_ · 2026-10-27