MiniMax · video model · Released Jul 31, 2026MiniMax H3
MiniMax general-purpose video model with multimodal references and synchronized stereo audio. Open H3-Base checkpoints; Context-IR and 2K regeneration are not included in the initial weights release.
- Arena score
- 1,495#1 of 48 · image to video
- Starting price
- —Not published
- Longest clip
- —Not published
- Resolution
- —With native audio
How it rates and what it costs
- Text to video arena
- 1,460 ±9#8 of 48 · 11,417 votes
- Image to video arena
- 1,495 ±5#1 of 48 · 57,112 votes
Arena scores come from blind, side-by-side human votes on the same prompts (higher is better; ± is the 95% range). Boards updated Sep 22, 2026.
The details
- Lab
- MiniMax
- Type
- Video model
- Released
- Jul 31, 2026
- Inputs
- Text prompts, Reference images, Audio, Video
- Output
- Video with sound
- Can do
- Text to video, Image to video, Video editing, Audio input, Native audio
- Availability
- 1 API host
About MiniMax H3
How good is MiniMax H3?
In blind side-by-side votes, MiniMax H3 scores 1,460 for text to video (#8 of 48) and 1,495 for image to video (#1 of 48). Higher is better; scores within about 18 points of each other are effectively tied.
Who makes MiniMax H3?
MiniMax H3 is a video model from MiniMax, released Jul 31, 2026.
What can MiniMax H3 do?
MiniMax H3 accepts text prompts, reference images, audio and video and produces video with sound. In practice: text to video, image to video, video editing, audio input and native audio.
Is MiniMax H3 open source?
Yes. MiniMax publishes the weights under the MiniMax H3 Community License Agreement, so you can run it on your own hardware.
Where can I use MiniMax H3?
It is listed by 1 API host we track.
What are the alternatives to MiniMax H3?
The closest video models from other labs, by release date, are Seedance 2.5 (ByteDance Seed, Jul 31, 2026), FLUX 3 Video (Black Forest Labs, Aug 4, 2026), Wan v3.0 Video (Alibaba (Qwen), Aug 23, 2026) and Grok Imagine Video 1.5 (xAI, May 30, 2026).