The author ported the 1.3B‑parameter LingBot‑World‑V2 causal‑fast video world model to Apple Silicon, running it on an M4 Pro Mac using PyTorch MPS instead of CUDA. The model generated 6.8 seconds of 832×464 video (121 frames requested, 109 written) with an MLX 8‑bit text encoder, completing in ~14.4 minutes. Peak memory usage reached ~16 GB of unified memory, demonstrating that the model can run on a single Mac without multi‑GPU or FlashAttention dependencies.