Maker-reported (not independently measured by JevMade): 49 layers over 4 steps · 6:07 to 3:34 · 41.7% reported reduction
- Primitives
- Not stated
- Platform
- RTX 4070
- Added
Source checked 2026-09-25 — opened the primary source directly.
Benchmarks & research
Let Jev select sparse-attention rates across MiniMax H3 layers to shorten video generation on an RTX 4070.
Screenshot unavailable. Open the experiment ↗
Maker-reported (not independently measured by JevMade): 49 layers over 4 steps · 6:07 to 3:34 · 41.7% reported reduction
Source checked 2026-09-25 — opened the primary source directly.