MiniMax
MiniMax

H3

Release Date 2026-08-03Proprietary ModelAPI available
Video overall3.0Top 25%
Text-to-video3.0Top 42%
Image-to-video3.0Top 25%
Video editing3.0Top 42%

MiniMax H3 is a multimodal model family that generates video with audio from text and visual or audio context. MiniMax presents it as a general generation system, but the released base weights should be distinguished from the complete hosted workflow. Public checkpoints cover frame-guided and reference-guided generation; context preprocessing and high-resolution regeneration were not included in the announced weight release and require separate services or implementations. AIB currently lacks enough comparison data to assign this model a performance band, so superiority over other video models is not established. Review the community license and deployment components, compare results using consistent prompts and references, and inspect character continuity, motion and sound alignment against production needs and processing cost.

API cost

$0.13/ sec

Output Speed

Very slow
247s
7-day median of 96 requests
SourceOpenRouterLast updated

Expert Performance Evaluation

1493112711121378

More about Arena Image-to-Video

See each evaluation’s detail page for its source and last updateData corrections & removal requests