Ten seconds of video in about eleven seconds: fal and Black Forest Labs launch FLUX 3 Max
AI platform fal has released the video model FLUX 3 Max together with Freiburg-based start-up Black Forest Labs. It is meant to match the best models while running considerably faster.
Photo: Benoît Prieur / Wikimedia Commons, CC0 (cropped and resized)
Video: YouTube · fal · youtube.com
AI provider fal on Monday unveiled FLUX 3 Max, a model for generating video from text or still images. It is based on the FLUX 3 foundation model from Freiburg-based start-up Black Forest Labs. fal’s research team post-trained it together with the developers, including with reinforcement learning, and optimised it for faster inference.
According to fal’s own measurements, a ten-second video is generated in 10.95 seconds; the rival model Seedance 2.5 needs 277.8 seconds and the original FLUX 3 just over 126 seconds. On its product page, however, fal states 12.45 seconds. In preliminary user preference tests by Artificial Analysis, FLUX 3 Max scored 1,139 Elo points, placing it among the top image-to-video models.
At launch, one second of 720p video costs 0.0425 US dollars, 75 per cent below the list price of 0.17 dollars; fal says the discount runs until 7 November. Black Forest Labs CEO Robin Rombach said FLUX 3 was built as a foundation for others to build on. fal presents the model in a video on its official channel.