mlx-dspark: Native Apple Silicon speculative decoding for DSpark and DFlash via MLX
Summary
mlx-dspark is a tool that runs DeepSeek's DSpark and z-lab's DFlash speculative-decoding drafters natively on Apple Silicon via MLX. These are lossless drafters (same output, just faster) designed to accelerate inference for the Qwen3 and Gemma-4 model families, plus any compatible DSpark/DFlash checkpoint. The tool supports CLI usage, Python integration, and serving an OpenAI-compatible API for use with LM Studio or other local tools.
Source
Twitter / Xmlx-dspark: Native Apple Silicon speculative decoding for DSpark and DFlash via MLXgithub.comKey quotes
· 3 pulledDeepSeek's DSpark and z-lab's DFlash speculative decoding — native on Apple Silicon via MLX.
Lossless drafters (same output, just faster) for the Qwen3 and Gemma-4 families — plus any matched DSpark / DFlash checkpoint.
Run them at the CLI, from Python, or serve an OpenAI-compatible API to LM Studio / any local tool.
You might also wanna read
DeepSeek open sources DSpark framework to accelerate LLM inference by up to 85%
Chinese AI company DeepSeek has open-sourced DSpark, a new MIT-licensed framework designed to accelerate LLM inference by up to 85%. DSpark
venturebeat.com·1d agoRunning a 284B AI Model on Your Desk: Our Real-World DSpark Deployment Log
DSpark: A Speculative Decoding Framework Using Semi-Autoregressive Generation and Confidence-Scheduled Verification for LLM Inference Acceleration
DSpark is a speculative decoding framework for Large Language Models (LLMs) that improves inference efficiency by combining semi-autoregress
arxiv.org·16h ago
DeepSpec/DSpark_paper.pdf at main · deepseek-ai/DeepSpec
github.com·11d ago
Optimizing LLM Inference by Combining NVIDIA DGX Spark and Apple Mac Studio Architectures
The article explores combining NVIDIA DGX Spark AI supercomputers with Apple Mac Studio systems to optimize large language model (LLM) infer
Comments
Sign in to join the conversation.
No comments yet. Be the first.