Fal Ai

H3 Max is a text-to-video and image-to-video model from fal Research, post-trained on top of the open-weight MiniMax H3 base and served through fal's hosted API. It is aimed at developers and creative teams already building on fal's generative media cloud, which provides AI developer teams with API access to over 600 image, video and audio models through serverless GPU infrastructure . The launch matters because H3 Max lands at the top of two independent leaderboards while running faster than most competitors, giving teams a single endpoint that no longer forces a tradeoff between quality and latency. According to fal, Design Arena scores H3 Max at an Elo of 1,341 on its image-to-video board, ahead of the base model it was post-trained from and every other model listed , and Artificial Analysis ranks H3 Max first with audio, at an Elo of 1,201 with a 95% confidence interval of plus or minus 11 over 2,177 samples. At $3.60 per minute it is also the cheapest model in that board's top fifteen . A five-second clip renders in under three seconds with synchronized audio, and fal is offering signed-in users five free 15-second generations every 24 hours through its in-browser sandbox. The model was built by fal, the company founded in 2021 by Burkay Gur and Gorkem Yurtseven after their time at Coinbase and Amazon. On the technical side, fal says it introduced significant new data during post-training, tuning specifically for stronger prompt adherence and aesthetics, and spent a huge portion of its post-training compute on verifiable RL tasks. The architecture was designed around fal's in-house inference engine, which the team has spent the past four years optimizing for diffusion models . For teams already routing image and audio work through fal, H3 Max is a drop-in way to add a top-ranked video model to the same stack.