MiniMax H3 Max

Overview

MiniMax H3 Max is a post-trained edition of MiniMax H3, built by fal. MiniMax open-sourced H3 in August 2026; fal retrained those weights for stronger prompt adherence and better aesthetics, then paired the result with their own inference stack. What comes back is a 5 to 15 second clip at up to 768p and 24fps, with stereo audio generated alongside the picture.

Why the name is confusing

Max reads like a quality tier, the way Pro does elsewhere, and the model sits under MiniMax's name. Neither is quite right. This is a third party's tuned version of an open model, not a larger or higher-fidelity H3 released by MiniMax. Knowing that explains the specification that surprises people most: the resolution ceiling is lower than the model it derives from, not higher.

Where the ceiling comes from

H3 ships as three parts. A preprocessor reads the brief and rewrites it into a structured form. A base model renders that at 768 pixels on the short edge. A third module then regenerates the result at 2K, using the original brief again so fine detail and small text are recovered rather than guessed at. MiniMax open-sourced the middle part. So a post-trained variant inherits a native 768p generator with no 2K stage behind it, which is the trade this model makes: it gives up the high-resolution finish and spends everything on how quickly the 768p arrives.

What the speed changes

At half a minute a generation is a submission you walk away from. At a few seconds it becomes something you iterate on, and that shift matters more than the raw number. Twenty variations of an ad hook get judged as finished motion instead of as a storyboard. A shot gets blocked, watched, adjusted, and watched again inside a single sitting. Work that was never worth the wait, exploring a direction you only half believe in, becomes cheap enough to simply try.

The lower resolution is the honest cost of that. For a feed, a course, an internal cut, or any draft on the way to something else, 768p is enough. For a master, it is not, and the model it derives from is the natural next step.

MiniMax H3 Max on Morphic

MiniMax H3 Max sits in the video model picker alongside MiniMax H3, Seedance, Kling, and Veo. Write the shot as a brief, generate, and the take lands on the Canvas, where you can run the same brief through another model and compare them side by side. The habits this model rewards, writing the action across the clip rather than describing a frame, and directing sound as deliberately as picture, carry to every one of them.