MiniMax-H3视频模型现已在Hugging Face公开发布

# MiniMax AI Releases MiniMax-H3 on Hugging Face Open-weight model added to Hugging Face shortly after Alibaba Qwen release. MiniMax AI announced that MiniMax-H3 is now publicly available. The weights were posted on Hugging Face under the MiniMaxAI organization. Clément Delangue reposted the announcement from the MiniMax account. Simon Willison observed that the release occurred within hours of Alibaba's Qwen 3.8 Max. The linked Hugging Face page states the goal is to advance and democratize artificial intelligence through open source and open science. No model size, benchmarks, or license details appear in the posts shown. ## Reactions from ranked influencers Qwen 3.8 Max and MiniMax-H3 within hours of each other MiniMax-H3 Is Now Publicly Available [...] MiniMax-H3 Is Now Publicly Available MiniMax-H3 Is Now Publicly Available ## Combined views 57.8K 2 posts, first seen 13h ago ## Sentiment ### Positive Read Many users celebrated MiniMax's open-weight H3 release on Huggin

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 [...] Q: When was MiniMax H3 released? Announced and available through the API on July 31, 2026. The open-weights release is promised "in the coming days"; as of August 1, 2026 there is no Hugging Face

Previous video stacks split into text-to-video, image-to-video, first-and-last-frame, subject reference, motion reference, and video editing, each often a separate expert model. MiniMax H3 folds those into one pretraining paradigm where reference and editing relationships are expressed in natural language. MiniMax’s example prompt makes the point: reference the camera movement from Video 1, have the character in Image 2 sing, match the vocals to Audio 3. ## Is it deployable? Today: yes, through the API and no, on your own hardware. MiniMax launched H3 on July 31, 2026 with the model live in the platform API under the model ID `MiniMax-H3` and in the consumer Hailuo AI app. [...] Input limits worth designing around: Reference images: up to 9. Reference videos: up to 3 clips, 2–15 s each, ≤15 s total. Reference audio: up to 3 clips, and audio cannot be sent without an accompanying image or video. Mixed input caps at 12 files total. Prompt length ≤7,000 characters; request body ≤64 M

MiniMax has formally open-sourced MiniMax H3, its general-purpose multimodal video system — better known to many creators through the Hailuo product line. The release matters less as another “new model dropped” headline, and more as a shift in how commercial-grade audiovisual generation gets distributed: hosted APIs for the full stack, downloadable weights for the core generator, and a clearer split between what you can self-host today versus what still lives behind official services. This post summarizes the open release in plain language, cross-checked against MiniMax’s technical announcement and the public Hugging Face model materials, with a focus on what changes for creative production and product integration. MiniMax H3 launch visual — official MiniMax materials

level. And this card tells you in one go what exactly this HiLo H3 is. It's a lightweight open weights video generation model from MiniMax, and the word HiLo literally means sea snail in Chinese. What makes it interesting is the control it gives you. It handles text-to-video, image-to-video, reference-to-video, and even first and last frame generation, so you can drive it however you want. As soon as they release it on Hugging Face, I'm going to make another video. We will install it locally, and we will test it out. This is, by the way, Fahad Mirza, and I welcome you to the channel. If you want to help out the channel, please become a member. I recently also have started this free weekly AI newsletter, which you can subscribe at fahadmirza.substack.com. Coming back to this model, you can [...] MiniMax just dropped HiLo 3, which is their new open weights video model, and today we are generating video with it hands-on. I'm going to use their API, though this model will be available on H

00:00 253K user avatar MiniMax (official) @MiniMax\_AI 17h This is how much you can do with a single RTX 5090 NOW with MiniMax H3. WE HAVE CROSSED A LINE.✊ user avatar Ryan Lightbourn @ryanlightbourn Aug 3 Generated locally with H3 on an RTX 5090 in a few minutes, from nothing but a text prompt. We have crossed a line. 00:00 88K user avatar Arena.ai @arena 16h Big news: MiniMax-H3 by @MiniMax\_AI is now the #1 open model in Video Arena: across both Text-to-Video and Image-to-Video. This is +280pts over the next best open model, hunyuan-video-1.5, and a huge improvement from Hailuo-2.3 at #27 (1199 pts) and Hailuo-02-pro at #28 (1197 00:41 user avatar MiniMax (official) @MiniMax\_AI Aug 3 MiniMax-H3 Is Now Publicly Available huggingface.co/MiniMaxAI/Mini… [...] ## Related Trending Stories on X MiniMax H3 AI Video Model Launches with 2K and Audio Features Other · 11.7K posts MiniMax Launches H3 AI Video Model with Audio and 2K Clips News · 34.2K posts