MiniMax H3 Turbo LoRA (larryvrh)
by larryvrh · Model · Video
Turbo LoRA for MiniMax H3 cuts sampling from 20 steps to about 6; ~780 MB file, Apache-2.0 per model card
AccessOpen weightsNeeds~50 GBDownload0.8 GBLicenseApache-2.0
Evidence
- Aug 23VRAM peaked just over 50 GB while sampling with the LoRA in ComfyUI; under 48 GB needs the earlier video's settings
Cost to run
Alternatives
Made from