One of the benefits of open source and open weight models is the flexibility they give to developers to come up with their own solutions. Take fals’ H3 Max: it is a post-trained video model that can generate a 5-second 720p video under 3 seconds. It ranks #1 against 12 leaving models as far as overall quality, prompt understanding and aesthetics. It is post-trained by fal on top of the open-weight base MiniMax H3 model. Here is how this was pulled off:
H3 Max is post-trained by fal on top of the open-weight base MiniMax H3 model. We introduced significant new data during post-training, tuning specifically for stronger prompt adherence and aesthetics. And spent a huge portion of our post-training compute on verifiable RL tasks. We also designed the architecture of H3 Max around fal’s in-house inference engine where we spent the past 4 years optimizing diffusion models with it. We prioritized quality first, then pushed speed as far as we could without compromising the aesthetics and prompt understanding. The result is a model that runs significantly faster without the usual tradeoff in quality.
If you check the website, you can see a 5-sec video costs $0.04 per second at 768p. This is 50% off, so you will pay about $0.08 per second in the future.
Introducing H3 Max, new post-trained video model by fal Research.
H3 Max ranks #1 for overall quality, prompt understanding, and aesthetics against leading video models, on both first-party and third-party independent evaluations while generating a 5-second 720p video under 3… pic.twitter.com/lSKkBRVXDR
— fal (@fal) August 26, 2026
