MiniMax’s H3 Max, launched just a week ago, gets a speed boost: its Turbo version is 2.5 times faster, with a 50% price cut.

Beating AI Express: AI inference platform fal has launched a faster, cheaper iteration of the H3 Max, which was released just one week ago. The H3 Max Turbo is built on MiniMax H3 post-training, optimized exclusively for fal’s proprietary inference system. In official tests of 5-second, 768p videos, generation latency fell from 3.49 seconds for the original H3 Max to 1.40 seconds—roughly a 2.5x speedup. While speed improved, quality saw a minor decline: fal’s internal testing benchmarks the original H3 Max at 100% quality, with the Turbo version scoring 97%. The company states that prompt comprehension, aesthetics, and overall quality are largely preserved, and the Turbo still outperforms the original MiniMax H3 and other tested video models. The H3 Max launched only on August 27, so fal is already delivering a faster, lower-priced variant just one week later. The regular price for the Turbo’s 768p offering is $0.04 per second, half of the H3 Max’s $0.08 per second. A launch promotion is now running with a 75% discount, bringing the rate to $0.01 per second—making a 1-minute video cost just $0.60.