# A 33-Billion-Parameter Video and Audio Model Now Runs on a MacBook, Slowly

Salvatore Sanfilippo has written a native Metal inference engine for MiniMax-H3, while a separate MLX port needs about 115 gigabytes of weights and takes just under 45 minutes per clip on an M5 Max.

- Published: 2026-08-11T06:26:24.055Z
- Canonical: https://polylog.news/ai/2026-08-11/a-33-billion-parameter-video-and-audio-model-now-runs-on-a-m
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [h3.c (GitHub)](https://github.com/antirez/h3.c), [Glonce](https://glonce.com/minimax-h3-video-model-ported-to-mlx-runs/), [Meta AI](https://ai.meta.com/blog/introducing-muse-image-muse-video-msl/)

MiniMax-H3 is a 33-billion-parameter joint video and audio diffusion model, and two independent efforts have now brought it to Apple Silicon without using PyTorch. Salvatore Sanfilippo, the creator of Redis, is building h3.c, a native infer…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-08-11/a-33-billion-parameter-video-and-audio-model-now-runs-on-a-m (subscription information: https://polylog.news/pricing).