How much does H3 Max cost?

H3 Max is billed per second of video on fal. The launch rate is $0.04 per second at 768p and $0.025 per second at 480p, discounted 50% through September 1. After the promotion the standard rates are $0.08 per second at 768p and $0.05 per second at 480p, so a 15-second 768p clip costs $1.20.

Can I try H3 Max for free?

Yes. fal gives five 5-second 768p generations with synchronized audio every day with no account and no sign-up, on a rolling 24-hour window. Signing in to the fal sandbox adds five more generations a day at up to 15 seconds each, which is enough to compare H3 Max against other video models before paying for anything.

How is H3 Max different from the base MiniMax H3?

H3 Max keeps MiniMax H3's architecture but adds a post-training pass by fal focused on prompt adherence and visual quality, plus a rebuilt serving path. fal measures roughly 35 times the throughput of the official MiniMax H3 endpoint, and Artificial Analysis ranks H3 Max ahead of the base model on both its image-to-video and text-to-video boards with audio.

Are the H3 Max weights available to download?

fal serves H3 Max through its own API, Playground and fal Agent rather than as a download. The base it was post-trained from, MiniMax H3, is public on Hugging Face under the MiniMax H3 Community License as a 33B model, so teams that need local weights can run the base and lose the fal post-training and speed work.

What aspect ratios and lengths does H3 Max support?

H3 Max generates clips of 5 to 15 seconds at 480p or 768p, with 768p at 1344x768 and 24 frames per second as the default. Text-to-video accepts 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16. Image-to-video also handles first-to-last keyframes through an optional end image.

fal · 2026-08-27 · major

H3 Max — fal's post-trained MiniMax H3 makes a 5-second clip in under 3 seconds

H3 Max is a video model that fal post-trained on the open-weights MiniMax H3. It returns a 5-second clip with synchronized audio in under 3 seconds and ranks first for image-to-video with audio on Artificial Analysis.

fal announcement artwork for the H3 Max video model

A post-trained MiniMax H3 that keeps the quality and returns about 35 times more clips per second.

Quick facts

Maker	fal Research
Base model	MiniMax H3 (open weights)
Output	480p or 768p, 5–15s, with audio
Speed	5-second clip in under 3 seconds
Ranking	#1 image-to-video with audio, Elo 1202 (Artificial Analysis)
Availability	fal API, Playground and fal Agent

Pricing

768p (launch) · 50% off through September 1	$0.04 / second
768p (standard)	$0.08 / second
480p (launch) · 50% off through September 1	$0.025 / second
480p (standard)	$0.05 / second

source ↗

What is it?

H3 Max is fal Research's post-trained version of MiniMax H3, the open-weights video model MiniMax published in August. fal added substantial new training data aimed at prompt adherence and visual quality, then served the result on its own inference stack. Output is 480p or 768p video, 5 to 15 seconds long, with synchronized audio produced in the same pass.

How does it work?

Post-training starts from the public MiniMax H3 base weights instead of a fresh architecture, so H3 Max inherits the base model's single-pass video-plus-audio generation. fal's inference team then rewrote the serving path: a 5-second clip comes back in under 3 seconds, which fal measures as roughly 35 times the throughput of the official MiniMax H3 endpoint and, on average, 15 times faster than models of comparable quality.

Why does it matter?

Waiting a minute per take is what makes video models awkward to build products on. At under 3 seconds a clip, H3 Max is fast enough to sit inside an interactive editor or an agent loop, and Artificial Analysis currently places it first among image-to-video models with audio at an Elo of 1202 — so the speed is not bought with quality.

Who is it for?

video creators and app developers

Frequently asked questions

How much does H3 Max cost?: H3 Max is billed per second of video on fal. The launch rate is $0.04 per second at 768p and $0.025 per second at 480p, discounted 50% through September 1. After the promotion the standard rates are $0.08 per second at 768p and $0.05 per second at 480p, so a 15-second 768p clip costs $1.20.
Can I try H3 Max for free?: Yes. fal gives five 5-second 768p generations with synchronized audio every day with no account and no sign-up, on a rolling 24-hour window. Signing in to the fal sandbox adds five more generations a day at up to 15 seconds each, which is enough to compare H3 Max against other video models before paying for anything.
How is H3 Max different from the base MiniMax H3?: H3 Max keeps MiniMax H3's architecture but adds a post-training pass by fal focused on prompt adherence and visual quality, plus a rebuilt serving path. fal measures roughly 35 times the throughput of the official MiniMax H3 endpoint, and Artificial Analysis ranks H3 Max ahead of the base model on both its image-to-video and text-to-video boards with audio.
Are the H3 Max weights available to download?: fal serves H3 Max through its own API, Playground and fal Agent rather than as a download. The base it was post-trained from, MiniMax H3, is public on Hugging Face under the MiniMax H3 Community License as a 33B model, so teams that need local weights can run the base and lose the fal post-training and speed work.
What aspect ratios and lengths does H3 Max support?: H3 Max generates clips of 5 to 15 seconds at 480p or 768p, with 768p at 1344x768 and 24 frames per second as the default. Text-to-video accepts 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16. Image-to-video also handles first-to-last keyframes through an optional end image.

Try it

https://fal.ai/models/minimax/h3-max/text-to-video