MiniMax H3 Max released just a week ago gets a speed boost: Turbo version is nearly 2.5 times faster, price cut in half

Percept Beating AI News Flash: The AI inference platform fal has just released an even faster and more cost-effective version of the H3 Max, just one week after its initial launch. The H3 Max Turbo is still based on the MiniMax H3 architecture and has been optimized for fal's in-house inference system. In official tests on a 5-second, 768p video, the processing delay has been reduced from 3.49 seconds on the H3 Max to 1.40 seconds on the Turbo version, an improvement of nearly 2.5 times.

While the speed has increased, there has been a slight decrease in quality. In fal's internal tests, the original H3 Max was rated at 100% quality, while the Turbo version scored 97%. The company stated that aspects such as keyword understanding, aesthetics, and overall quality have been mostly preserved, and that the Turbo version still outperforms the original MiniMax H3 and other benchmarked video models.

The H3 Max itself was only launched on August 27th, and just a week later, fal has already accelerated and reduced the price once again. The regular price for Turbo at 768p is $0.04 per second, only half of the H3 Max's regular price of $0.08 per second. As a special launch promotion, there is an additional 75% discount, bringing the price down to just $0.01 per second, making a one-minute generated video cost only $0.6.