MiniMax plans to release H3 2K model and sparse-attention code
MiniMax says it plans to release H3's separate 2K regeneration model and sparse-attention reference code, giving developers more of its video-generation stack to inspect and test locally.
By Ryan Merket · Published
Primary source: MiniMax
Why it matters
MiniMax's downloadable H3 release initially produces 768p output, while the separate 2K regeneration stage remains part of its hosted workflow. Releasing that stage and sparse-attention code would let engineers test H3's higher-resolution process and potential inference savings on their own hardware.
MiniMax founders Yan Junjie and Yeyi Yun are preparing to expand the downloadable parts of H3, the Shanghai-based AI developer's multimodal video model. In an August 7th, 2026 community discussion, H3 team members said MiniMax plans to release its 2K regeneration stage, publish sparse-attention reference code and develop a unified model for image generation and editing.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.