Unreleased MiniMax H3 Surfaces with Native Multimodality and 12 Mixed Inputs

bennash · x · 2026-07-31

A developer is testing the unreleased MiniMax H3 model from Hailuo AI. The model boasts native multimodal understanding, allowing users to mix and match up to 12 references across video, text, image, and audio inputs. It is being positioned as a potential competitor to Seedance 2.0.

Related event: MiniMax H3 Model Early Tests Reveal Native Multimodal and 2K Video Capabilities(5 posts)→

Original post →

More from Models

Models channel →