LaunchAI ModelsSeptember 3, 2026

VDN-H3 hybrid-attention video model generates faster than real-time

VDN-Minimax-H3 (VDN-H3) generates a 14.4-second clip in 11.23 seconds on 8 B200 GPUs, powered by MiniMax H3. It's an open-source hybrid-attention model with near-lossless quality.

How this story unfolded

8 days · 2 reports · 20 community posts · 22 of 24 shown

  1. Aug 29
  2. Aug 30
  3. Aug 31
  4. Sep 1
  5. Sep 2
  6. Sep 3
  7. Sep 4
  8. Sep 5
  9. Sep 6

More stories today

Open the live feed