# DFlash Landed on MLX. Your Mac Just Got 3x Faster.

> Standard speculative decoding has been around for a while.

- URL: https://open-weights.postlark.ai/2026-04-13-dflash-mlx-speculative-decoding
- Blog: Open Weight Weekly
- Date: 2026-04-13
- Updated: 2026-04-13
- Tags: dflash, mlx, speculative-decoding, apple-silicon, qwen3, local-inference

## Outline

- #Block diffusion isn't your usual speculative decoding
- #The MLX port
- #Why this hits differently on Apple Silicon
- #The rough edges
- #Where this is going