Overview
FLUX 3 Max is a video model made by fal Research together with Black Forest Labs, built on the FLUX 3 base. The split is in the name: BFL supplies the foundation model, fal handles post-training, inference optimisation and systems engineering, and the resulting weights serve both text-to-video and image-to-video at 720p.
The pitch is holding speed and quality at the same time. On Artificial Analysis's preliminary image-to-video arena it scores 1,139 Elo, tied for first with overlapping confidence intervals, and it returns a 10-second 16:9 720p clip in 12.45 seconds. fal says that is around 22 times faster than Seedance 2.5, which in practice means you no longer need a fast model for drafts and a slow one for finals, since the take you iterate on is the take you ship.
Billing is per second of output video with no subscription and no minimum. That is 0.0425 dollars per second at 720p against a list price of 0.17, with the 75 percent discount running until November 7. The API exposes text-to-video, image-to-video and a keyframe-to-video endpoint, with Python and JavaScript SDKs.
Key Features
- Near real-time output: A 10-second 720p clip returns in about 12.45 seconds, which makes twenty takes of a shot practical.
- Tied for first on image-to-video: Scores 1,139 Elo on the preliminary Artificial Analysis arena, tied with the leader within the same confidence interval.
- One model, two entry points: Text-to-video and image-to-video share the same post-trained weights, so quality holds whichever way you start.
- Hold shape at 720p: Fast-moving subjects keep their form from first frame to last, from a horse's stride to the grain of painted wood.
- Keyframe to video: A third endpoint takes start and end frames, useful for pinning down camera moves and transitions.
- Per-second billing: Only output seconds are charged, with no monthly plan and no minimum, so trial costs scale by the clip.
Use Cases
- Ad and ecommerce teams draft dozens of storyboards, then ship the chosen take as final.
- Product stills become motion showcases without a photo shoot.
- Film pre-production tests camera placement, lighting and moves quickly.
- Games produce concept trailers and compare several versions from one prompt.
- Social video batches are costed per clip thanks to per-second pricing.
- Start and end frames pin down transitions between shots.
Pros
- Output speed approaches real time, allowing many more iterations.
- Image-to-video quality sits level with the top of the arena, so drafts and finals can be the same version.
- Text and image entry points share weights, keeping quality consistent.
- Per-second pricing means no subscription commitment.
- It sits alongside Seedance, Kling and Veo on fal, so it can be chained in one workflow.
- Python and JavaScript SDKs keep integration to a few lines.
Pricing
Billed per second of output video: 0.0425 dollars per second at 720p, down from a list price of 0.17 dollars, with the 75 percent discount running to November 7. There is no minimum spend and no subscription, and text-to-video and image-to-video cost the same. Access uses a fal API key with the Python or JavaScript SDK over an asynchronous submit-and-poll flow, and prompts can also be tried in ChatGPT or the fal Agent playground.
Summary
Video models used to carry a hidden tax: choose speed and accept mush, choose stability and wait minutes, so production split into a draft model and a final model. FLUX 3 Max tries to remove that split by post-training one model to return 10-second clips in about 12 seconds while staying near the top of the quality board. For teams shipping dozens of clips a day, the improvement is not in parameters but in making the number of attempts affordable.
Version History
- FLUX 3 Max listed: 10-second 720p video in 12 seconds (2026-10-31): Form at listing: an image-to-video and text-to-video model post-trained by fal Research with Black Forest Labs on the FLUX 3 base. Scores 1,139 Elo on the preliminary Artificial Analysis image-to-video arena, tied for first, and returns a 10-second 16:9 720p clip in about 12.45 seconds. Text-to-video, image-to-video and keyframe-to-video endpoints, Python and JavaScript SDKs. Billed per output second at 0.0425 dollars for 720p against a 0.17 list price, with a 75 percent discount through November 7.