News
Another domestic model is open sourced! No. 1 in the world in audio video editing, 16 chips and platforms adapted on the first day
2 min read
Source: zhidx.com
Zhidongxi Author | Yang Jingli Editor | Li Shuiqing Zhidongzhi reported on August 3 that today, MiniMax officially open sourced the new generation universal video model MiniMax H3. MiniMax H3 is a universal full-modal generation system that can uniformly understand multi-modal contexts composed of text, images, videos and audio, and generate videos of up to 15 seconds, up to 2K resolution and with native stereo audio. Previously on July 31, MiniMax H3 was released. Currently, in the Artificial Analysis audio video editing list, MiniMax H3 ranks first with an Elo score of 1130 points, leading domestic and foreign video models such as Gemini Omni Flash, HappyHorse-1.0, and Wan 2.7. ▲MiniMax H3 ranks first on the audio video editing list. The H3 system consists of three modules: H3-Context-IR, H3-Base, and H3-Regenerate-2K. Developers can download the MiniMax H3 model directly from Hugging Face. H3-Base currently supports deployment through inference frameworks and workflows such as SGLang, vLLM, diffusers and ComfyUI. In sync with the model open source, domestic and foreign chip manufacturers such as Huawei Ascend, Moore Thread, Muxi, Haiguang Information, Kunlun Core, Tianshu Intelligent Core, Biren Technology, AMD, and Intel, as well as development communities and cloud inference platforms such as Hugging Face, Moda ModelScope, ComfyUI, RunningHub, fal, and inference frameworks such as vLLM-Omni and SGLang, a total of 16 ecological partners have completed relevant adaptation support. Open source address: huggingface.co/MiniMaxAI/MiniMax-H3 Model experience address: H3-2K direct outlet: platform.minimax