What Is Flux 3 Max? Features, Pricing and How to Use It
Flux 3 Max is fal's fast video model built on Black Forest Labs' FLUX 3. Modes, clip length, resolution, audio, launch pricing to 7 Nov 2026, benchmarks and limits.
Flux 3 Max is an AI video model that turns a photo, a pair of keyframes or a text prompt into a 5 to 20 second clip with sound. fal released it on 19 October 2026. fal Research built it with Black Forest Labs, starting from Black Forest Labs' FLUX 3 foundation model and adding post-training, reinforcement learning and inference optimization (fal press release). Despite the FLUX name, which many people know from image generation, Flux 3 Max is a video model.
Its selling point is speed at top-tier quality. fal reports about 11 seconds to render a 10-second clip, and the model ties for first place on the Artificial Analysis image-to-video leaderboard. This guide covers what it does, what it costs, how it compares and how to use it. You can try it now in the Flux 3 Max video generator on our homepage.
Key Takeaways
- Flux 3 Max is a video model from fal, post-trained with Black Forest Labs on FLUX 3, released 19 October 2026.
- It has three modes: text-to-video, image-to-video and keyframes-to-video. Clips run 5–20 seconds at 720p, with optional 1080p and native audio.
- fal measured 10.95 seconds to generate a 10-second clip. It scores 1,138 Elo on the Artificial Analysis image-to-video board, tied with Seedance 2.5 and MiniMax H3.
- fal's list price is $0.17 per second at 720p. A 75% launch discount runs through 7 November 2026.
- It does not edit or extend existing videos, output 4K, or lock a face from reference photos.
Where Did Flux 3 Max Come From?
Flux 3 Max is the second step in Black Forest Labs' move into video. Black Forest Labs announced FLUX 3, its first video model, on 23 July 2026, with native audio and clips up to 20 seconds (The Decoder). Three months later, fal released a post-trained version under the Max name.
| Date | Event |
|---|---|
| 23 Jul 2026 | Black Forest Labs announces FLUX 3, its first video model, with native audio |
| 19 Oct 2026 | fal releases Flux 3 Max with text-to-video, image-to-video and keyframes-to-video |
| 19 Oct 2026 | Black Forest Labs posts "Introducing FLUX 3 Max" on X (source) |
| 7 Nov 2026 | fal's 75% launch discount ends |
Black Forest Labs CEO Robin Rombach framed the release as the point of FLUX 3: "We built FLUX 3 as a foundation for others to build on," he said in fal's release. fal calls Max the "first iteration" and says it will keep post-training the model.
What Can Flux 3 Max Do?
Flux 3 Max makes short videos with sound from three kinds of input, through one API on fal.
- Image-to-video. Upload a still and describe the motion. The image becomes the first frame. Try it in our image-to-video tool.
- Text-to-video. Describe the scene, action and camera in words. Try it in our text-to-video tool.
- Keyframes-to-video. Give a first frame and a last frame, and the model fills the motion between them. On our homepage this is the First & Last Frame tab.
| Spec | Flux 3 Max |
|---|---|
| Modes | Text-to-video, image-to-video, keyframes-to-video |
| Clip length | 5–20 seconds |
| Resolution | 720p, optional 1080p upscale |
| Aspect ratios | Landscape, portrait and square |
| Audio | Generated with the video; can be switched off |
| Reasoning setting | Low or high on text-to-video; high is slower |
| Where it runs | fal (not on Black Forest Labs' own API as of 20 Oct 2026) |
Sources: fal press release, VentureBeat (19 Oct 2026), OrcaRouter (20 Oct 2026).
How Fast and How Good Is Flux 3 Max?
Flux 3 Max is fast enough to iterate on almost in real time, and it scores with the best video models on blind votes.
Speed. In fal's latency tests, a 10-second Flux 3 Max video took 10.95 seconds, against 126.44 seconds for FLUX 3 and 277.80 seconds for Seedance 2.5. Black Forest Labs rounds it to "a 10-second video in 12 seconds." An independent test by SunoMV on fal's public endpoints, queue included, measured 13–21 seconds for a 10-second text-to-video clip and 16 seconds for a 5-second image-to-video clip (SunoMV, 20 Oct 2026).
Quality. On the Artificial Analysis image-to-video leaderboard, retrieved 20 October 2026, Flux 3 Max (720p) scores 1,138 Elo across 5,054 blind votes. That ties it with Seedance 2.5 (1,138) and MiniMax H3 (1,140), whose confidence intervals overlap with it. It is 52 points above FLUX 3.
Both sets of numbers carry caveats. The speed figures are fal's own, and no one has independently repeated the Seedance comparison. The leaderboard covers image-to-video only, and fal describes the Elo result as preliminary. For the full breakdown, read Flux 3 Max vs Seedance 2.5 and the best image-to-video models of October 2026.
How Much Does Flux 3 Max Cost?
On fal, Flux 3 Max lists at $0.17 per second at 720p and $0.29 per second at 1080p, the same as FLUX 3 (SunoMV). fal is taking 75% off for the first three weeks, through 7 November 2026, which brings 720p to $0.0425 per second (AlphaSignal).
| Clip | Launch rate (to 7 Nov 2026) | List rate |
|---|---|---|
| 5 s at 720p | $0.21 | $0.85 |
| 10 s at 720p | $0.43 | $1.70 |
| 20 s at 720p | $0.85 | $3.40 |
| 10 s at 1080p | $0.73 | $2.90 |
Our arithmetic from fal's per-second rates, rounded to the cent. Prices on other hosts differ.
You do not need a fal account to use it. On flux3max.top you sign in with Google and pay in credits, and the launch discount is passed through until 7 November. Current credit costs for each length and resolution are on our pricing page.
How to Use Flux 3 Max
You can make a Flux 3 Max video in four steps on our homepage, with no API key.
- Pick a mode. Choose Image to Video, Text to Video, or First & Last Frame.
- Add your input. Upload a photo, upload a first and last frame, or skip the upload for text-only.
- Write the prompt. Describe the subject, the action, the camera move and the sound. Then pick the length, resolution and whether to generate audio.
- Generate. Turn on Preview mode to block out the shot with a quick FLUX 3 draft, then switch it off to render the final Flux 3 Max clip.
A prompt structure that works
Long clips go better when the prompt says what happens when. The template we use on our homepage splits a 10-second clip into timed beats:
Cinematic western, 35mm film look. 0-3s: wide shot of a lone rider crossing a dusty main street at golden hour. 3-7s: slow push-in as he stops and tips his hat, wind lifting the dust. 7-10s: close-up of his eyes under the brim, distant church bell, horse snorting.
Three habits help:
- Name the camera move ("slow push-in", "dolly left", "locked-off wide shot"). SunoMV found Flux 3 Max on high reasoning followed a scripted camera move that FLUX 3 missed.
- Describe the sound. The model generates audio, so say what you want to hear.
- Use high reasoning for final text-to-video renders. In SunoMV's test, low reasoning was faster but shifted the color grade and changed how the character looked.
What Flux 3 Max Cannot Do Yet
Flux 3 Max is narrower than its base model. Plan around these gaps:
- No video editing or video-to-video. It makes new clips only. FLUX 3 can edit existing footage.
- No extension. It cannot continue a clip from its last frame. Chain clips, or use our AI video extender.
- No 2K or 4K. It tops out at 1080p.
- No face lock. It can pin a start and end frame, but it does not hold one person's face across separate clips from reference photos. For a recurring cast, a reference-driven model such as Seedance 2.5 works better. Our reference-to-video tool covers that job.
- 20 seconds per clip. Seedance 2.5 and Wan 3.0 go to 30.
Sources: SunoMV, VentureBeat. For how these trade-offs compare with the base model, see Flux 3 Max vs FLUX 3.
Frequently Asked Questions
Is Flux 3 Max an image model or a video model?
A video model. Earlier FLUX releases from Black Forest Labs were image models, but Flux 3 Max generates video with sound from images, keyframes or text.
Who made Flux 3 Max?
fal Research post-trained it in collaboration with Black Forest Labs, on top of Black Forest Labs' FLUX 3 foundation model. fal hosts it.
Is Flux 3 Max free?
No. It is a paid model priced per second of video. On fal, launch pricing is $0.0425 per second at 720p until 7 November 2026, then $0.17. On our site you pay with credits; see the pricing page.
How long can a Flux 3 Max video be?
5 to 20 seconds per clip.
Does Flux 3 Max make videos with sound?
Yes. It generates audio along with the video, and you can turn audio off.
Can I use Flux 3 Max without an API key?
Yes. Our Flux 3 Max video generator runs it in the browser after you sign in. Developers can call it directly on fal.
The Bottom Line
Flux 3 Max is a fast, top-ranked image-to-video model from fal and Black Forest Labs. It makes 5–20 second clips with sound in seconds rather than minutes, and it is heavily discounted until 7 November 2026. It is not the right tool for editing footage, 4K output or a recurring character, and its speed claims still need independent confirmation.
- Best for: b-roll, product shots, first-and-last-frame transitions, social clips, and fast storyboarding.
- Look elsewhere for: face-locked casts, clips over 20 seconds, video editing.
Upload a photo to the Flux 3 Max video generator and see how fast your first clip comes back.