O conteúdo está em inglês; a navegação e os filtros mantêm o idioma selecionado.

MiniMax H3 Model Weights Appear Public, With Omnimodal Generation Support

MiniMax has released H3 as a general-purpose omnimodal generation system for text, images, video, and audio. The official Hugging Face page is now available, although earlier reports said the weights would arrive within days.

Publicado em 06/08/2026, 09:20:34Última atualização 06/08/2026, 09:20:34Publicação editorial

MiniMax has released MiniMax H3, a general-purpose model designed to understand text, images, video, and audio within a unified context. According to the company, it can generate video with native stereo audio at up to 2K resolution and 15 seconds in length. The company positions it for advertising, e-commerce, branding, product design, UI/UX, and game-related workflows.

An official MiniMaxAI/MiniMax-H3 page is now available on Hugging Face. Its documentation says the initial public release supports inference with full attention, while the sparse-attention implementation intended to reduce the cost of long multimodal sequences will be released separately. H3 is also promoted through MiniMax’s API and consumer-facing platforms, subject to regional availability and applicable terms.

The evidence contains a timing discrepancy. Reuters and TechTimes described the weights as planned for release within days, while the current Hugging Face listing indicates that a model release is already accessible. The public-release conclusion is therefore well supported, but the exact launch timing and scope of the open license should be checked against the latest official model card and license text.

Evidências das fontes 6

A verificação de fontes não garante exatidão permanente. Considere a fonte original, a data e as correções posteriores.

What Is MiniMax H3 (Hailuo 3.0)? The Open-Weight ...huggingface.co · Complementar

TL;DR — MiniMax H3 (the official name of what most people call Hailuo 3.0) is the third generation of MiniMax's Hailuo video line, but it deliberately stops behaving like a video model. It reads text, images, video, and audio as one unified context, then generates a 4–15 s clip at 2K/24 fps with native stereo audio — no separate audio stage, no post-hoc upscaler. It supports up to 9 reference images, 3 reference video clips, and 3 reference audio clips per generation. On the Artificial Analysis leaderboards it ranks #1 in Video Editing, #2 in Text-to-Video, #3 in Image-to-Video. Weights are promised "in the coming days" under a planned MiniMax Community License (commercial use for organizations under $20M revenue, with attribution), but no Hugging Face model card exists yet as of August [...] MiniMax H3 is the official model name; Hailuo 3.0 — also written Hailuo 03 — is the widely used alias, after the Hailuo AI app it ships in. They are one model, the direct successor to Hailuo 2.3.