FastH3 Now Runs Locally on Apple Silicon and DGX Spark
Vandy_simp · reddit · 2026-09-03
The FastVideo team released local inference paths for FastH3: via MLX on Apple Silicon and on one or two NVIDIA DGX Sparks. Maintained recipes are exposed through Python, CLI tools, and a local OpenAI-compatible server/playground, covering setup, weight conversion, memory constraints, and reproducible generation. The team notes it is not yet a drag-and-drop ComfyUI workflow and is asking users whether native nodes, an API-backed node, or example workflows would be most useful. Code is open-sourced on GitHub.
More from Infra
- SGLang Summit sponsors grow from 11 to 36 in one week; Lip-Bu Tan, Lilian Weng to speak — BanghuaZ · 2026-10-18
- Yacine jokes his datacenter KPI is gigawatt-hours: get paged if not burning enough power — yacineMTB · 2026-10-18
- 1.68x inference speedup drops, directly accelerating GRPO post-training — yacineMTB · 2026-10-18
- OpenRouter is only a fraction of the total addressable market, poster argues — sloppenheimer · 2026-10-18
- Developer builds change alerts for LLM prices and retirements across Bedrock, Azure and Vertex — jsandhu0780 · 2026-10-18
- 12.3M tokens in 27 minutes for $0.14: agent session with 99.6% cache hit — teortaxesTex · 2026-10-18