88Z AI, 3 August — MiniMax has officially open-sourced its new general video model, MiniMax H3, which supports multimodal input and generates high-quality videos. The model consists of three modules, with the base module already open-sourced and supporting local deployment on mainstream frameworks.
At a glance
- MiniMax H3 supports multimodal input and generates 2K resolution, 15-second videos with 32kHz stereo audio.
- The model consists of three modules, with H3-Base open-sourced and supporting local deployment on SGLang and vLLM frameworks.
The update
MiniMax has officially launched and open-sourced MiniMax H3, a new generation general-purpose video model. This model features robust multimodal processing capabilities, accepting text, images, video, and audio as inputs. In terms of output quality, it can generate videos with up to 2K resolution, a duration of 15 seconds, and 32kHz stereo audio. The architecture of MiniMax H3 comprises three core modules: H3-Context-IR, H3-Base, and H3-Regenerate-2K. Currently, the H3-Base module has been open-sourced and is compatible with mainstream inference frameworks such as SGLang and vLLM, allowing developers to implement local deployment and applications based on these frameworks.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- MiniMax 稀宇科技Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.