Release
FastVideo releases FastH3 Preview v1, an open-weight 4-step distillation of MiniMax H3
On August 27, 2026, the FastVideo project released FastH3 Preview v1, which its repository describes as an open-weight, 4-step sparse-distilled MiniMax-H3 model for synchronized video and audio generation, developed with Nuva Lab and the NVIDIA FastGen team. The recommended checkpoint was trained with data-free DMD2 and VSA-H3 sparse attention, requires FastVideo's VSA-H3 attention backend, and is also offered as a matching LoRA adapter. Its model card says the checkpoint inherits the MiniMax H3 Community License. USASI has not tested FastH3.123
Related catalog entries
Sources
Share or cite this page
- X (opens X)
- Bluesky (opens Bluesky)
- LinkedIn (opens LinkedIn)
- Reddit (opens Reddit)
- Hacker News (opens Hacker News)
- Email (opens your email app)
- RSS updates
Cite this page
APA style
USASI. (2026, October 13). FastVideo releases FastH3 Preview v1, an open-weight 4-step distillation of MiniMax H3. United States of America Superintelligence. https://unitedstatesofamericasuperintelligence.com/news/fastvideo-releases-fasth3-preview-v1-open-weights/
BibTeX
@misc{usasi_news_fastvideo_releases_fasth3_preview_v1_open_weights,
author = {{USASI}},
title = {FastVideo releases FastH3 Preview v1, an open-weight 4-step distillation of MiniMax H3},
year = {2026},
month = oct,
howpublished = {\url{https://unitedstatesofamericasuperintelligence.com/news/fastvideo-releases-fasth3-preview-v1-open-weights/}},
note = {United States of America Superintelligence. Last updated 2026-10-13}
}The date is when this page was last updated. Add the date you read it if your style needs one.