FLUX 3 Max: the fast new video model, what it costs and how to use it
By GetStepSnap · Checked against fal's press release and model pages on October 22, 2026
FLUX 3 Max is a video model, and its selling point is speed. fal, the platform that runs it, says a 10-second clip takes about 11 seconds to make. It turns a photo or a sentence into 5 to 20 seconds of video with sound, at 720p or 1080p. fal Research trained it further with Black Forest Labs, the makers of FLUX, and fal released it on October 19, 2026.
It's cheap right now and won't stay that way. Until November 7, fal charges $0.0425 a second at 720p, 75% off. After that it's $0.17 a second, so a 10-second 720p clip that costs $0.43 today will cost $1.70.
Despite the name, it isn't an image model. For pictures, Black Forest Labs has FLUX 3 and FLUX 3 Klein, which you can use here.
FLUX 3 Max at a glance
| Made by | fal Research, post-trained with Black Forest Labs on their FLUX 3 model |
| Out | October 19, 2026, on fal.ai |
| Starts from | A sentence, a first frame (and an optional last frame), or several keyframes |
| Length | 5 to 20 seconds, in whole seconds |
| Size | 720p (the default) or 1080p |
| Shape | 21:9, 2:1, 16:9, 4:3, 1:1, 3:4 or 9:16, or follow the first frame |
| Sound | Made with the picture, on by default |
| Prompt help | It expands your prompt before it starts (reasoning high, the default); low is faster |
What a clip costs, now and after November 7
Billed per second of finished video. The launch discount runs from October 15 to November 7, 2026; the second price in each cell is what fal charges after that.
| Size | 5 seconds | 10 seconds | 20 seconds |
|---|---|---|---|
| 720p | $0.21after Nov 7: $0.85 | $0.43after Nov 7: $1.70 | $0.85after Nov 7: $3.40 |
| 1080p | $0.36after Nov 7: $1.45 | $0.73after Nov 7: $2.90 | $1.45after Nov 7: $5.80 |
How fast, and how good
fal's numbers: a 10-second video in 10.95 seconds, more than 22 times faster than Seedance 2.5 and 11.5 times faster than the original FLUX 3, which took 126.44 seconds. That is the model's own working time. What you wait for also includes the queue, the prompt rewrite (shorter with reasoning set to low) and the upload.
On quality, fal points to early votes on the Artificial Analysis Video Arena, where people pick the better of two clips without knowing which model made which: 1,139 Elo across 4,868 votes, among the top image-to-video models, and up from 1,089 for the original FLUX 3. These are early numbers reported by fal; we haven't run our own test yet.
FLUX 3 Max next to the other new video models
Price per second at 720p on fal.ai, and the longest clip in one go, on October 22, 2026.
| Model | 720p, a second | Longest clip |
|---|---|---|
| FLUX 3 Max | $0.0425 until Nov 7, then $0.17 | 20 seconds |
| Wan 3.0 | $0.10 | 30 seconds |
| Kling 4.0 | $0.157 (in fal's playground; API coming soon) | 30 seconds |
| Seedance 2.5 | $0.473 | 30 seconds |
Until November 7, FLUX 3 Max is the cheapest of the four by far. After that, Wan 3.0 costs less per second and goes longer; FLUX 3 Max is the one to pick when you need the clip fast.
Three ways to start a video
- From a sentence. Describe the scene; it picks a 16:9 frame unless you choose another shape.
- From a photo. Your image is the first frame, and you can add a last frame for it to land on. The video takes the photo's shape.
- From keyframes. Several images, each pinned to a moment, and it moves between them.
Writing a prompt for it
- From a photo, describe what happens next, not what the photo already shows. The first frame takes care of the look.
- Give dialogue in quotes and say who speaks. fal's own example has two characters at a desk, one line of dialogue and a reaction.
- Sound can have a direction. fal's example puts birdsong on the left and wind on the right.
- Leave reasoning on high to let the model expand a short prompt; set it to low when your prompt is already complete and you want the result sooner.
- Pick the length on purpose. It's billed by the second, and after November 7 each extra second at 720p costs $0.17.
What we're doing
GetStepSnap runs new AI models without a subscription: you buy credits once and use them on any tool. Today we make images. We plan to add video models here, and we'll price FLUX 3 Max on its full price after November 7, not the launch discount, so the price you see won't jump. Until then, the first frame is where a good video starts, and you can make one with our photo generator.
Make a first frameQuestions
What is FLUX 3 Max?
A video model that fal Research post-trained with Black Forest Labs on their FLUX 3 model. It makes 5 to 20 seconds of video with sound, at 720p or 1080p, from a sentence, a first frame or several keyframes. fal released it on October 19, 2026.
Is FLUX 3 Max an image model?
No, it makes videos. Black Forest Labs' FLUX 3 image models are different: FLUX 3 makes pictures from a sentence and FLUX 3 Klein edits a photo, and you can use both on GetStepSnap.
How much does FLUX 3 Max cost?
It's billed per second of video. Until November 7, 2026, fal charges $0.0425 a second at 720p and $0.0725 at 1080p, 75% off. After that it's $0.17 and $0.29 a second, so a 10-second 720p clip goes from $0.43 to $1.70.
How fast is FLUX 3 Max?
fal says it makes a 10-second video in 10.95 seconds, more than 22 times faster than Seedance 2.5 and 11.5 times faster than the original FLUX 3. Those are fal's own numbers; the time you wait also includes the queue, the prompt rewrite and the upload.
Is FLUX 3 Max good?
In early votes on the Artificial Analysis Video Arena, where people compare two clips without knowing which model made them, it scored 1,139 Elo across 4,868 votes, according to fal. That puts it among the top image-to-video models, up from 1,089 for the original FLUX 3.
How long can FLUX 3 Max videos be?
5 to 20 seconds, picked in whole seconds. Wan 3.0, Kling 4.0 and Seedance 2.5 go to 30 seconds.
Where can I use FLUX 3 Max?
Through fal.ai, where it launched: text to video, image to video and keyframes to video, from the playground or the API. fal's launch announcement names no other place to use it.
Sources, checked October 22, 2026: fal launches FLUX 3 Max (fal press release, October 19, 2026), FLUX 3 Max image to video on fal.ai, FLUX 3 Max text to video on fal.ai, FLUX 3 Max keyframes to video on fal.ai.
