FLUX 3 Max: What It Is, Speed, Price & Specs
FLUX 3 Max is fal Research's post-trained FLUX 3 video model: 5–20 s clips with sound, 720p or 1080p, $0.17/s list. Speed claims, Elo, price, where to use it.
Flux 3 Image Team
FLUX 3 Max is a real model, and it makes video, not images. fal Research post-trained it from Black Forest Labs' FLUX 3 video model, in collaboration with Black Forest Labs, and it went live on fal on October 16, 2026. It turns a text prompt, or a start frame with an optional end frame, into a 5 to 20 second clip with sound at 720p or 1080p. This guide sticks to what the makers and independent trackers have published, read on October 21, 2026, and says where they disagree.
TL;DR
- What it is: a video model made by fal Research in collaboration with Black Forest Labs, post-trained from FLUX 3 (Black Forest Labs' own video model, also called FLUX 3 Video). It is not an image model and not a subscription plan.
- What it makes: 5 to 20 second clips, in whole seconds, at 720p or 1080p, with sound generated alongside the picture (it can be switched off).
- Inputs: a text prompt, or 1 to 10 images: one sets the start frame, two set the start and end frames, and three to ten become keyframes spread through the clip.
- Speed: fal measured a 10-second clip in 12.45 seconds. Its "22× faster than Seedance 2.5" claim rests on its launch chart; a later fal chart narrows the gap to about 5.5×. Two independent tests found about 16 seconds per 5-second clip including the queue, and 20.5 seconds from submit to file for a 10-second clip.
- Quality: on Artificial Analysis's image-to-video leaderboard, FLUX 3 Max (720p) scores 1138 ±15 Elo, statistically tied with Seedance 2.5 (1138 ±17) and MiniMax H3 (1140 ±16).
- Price: $0.17 per second at 720p and $0.29 per second at 1080p list on fal, 75% off until November 7, 2026. On this site it costs 64 credits per second at 720p.
- Use it without code: fal's playground, Krea's FLUX 3 Max page, or the FLUX 3 Max generator on this site. We found no free route.
FLUX 3 Max at a glance
| FLUX 3 Max | |
|---|---|
| Made by | fal Research, in collaboration with Black Forest Labs |
| Base model | FLUX 3 (Black Forest Labs' video model) |
| Live on fal | October 16, 2026; launch coverage followed on October 19 |
| Media | Video with native audio |
| Modes | Text to video; image to video with a start frame and optional end frame; keyframes (up to 10) |
| Length | 5 to 20 seconds, in whole seconds |
| Resolution | 720p or 1080p |
| Aspect ratios | Auto, 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, 9:16 |
| Speed (fal's test) | 12.45 seconds for a 10-second 720p clip |
| Quality | 1138 ±15 Elo, image-to-video, Artificial Analysis (720p) |
| List price on fal | $0.17/s at 720p, $0.29/s at 1080p |
| Price here | 64 credits/s at 720p, 108 at 1080p |
What is FLUX 3 Max?
FLUX 3 Max is a post-trained video model. fal Research took Black Forest Labs' FLUX 3, the video model BFL released in its FLUX 3 family, and kept training it for closer prompt following and more polished video, then optimized it to run faster. fal's own FAQ puts the split this way: Black Forest Labs brought the foundation model; fal Research brought post-training, inference optimization and systems engineering. Black Forest Labs features it on its FLUX 3 model page under "Introducing FLUX 3 Max", with a link to try it on fal.
Two naming traps catch people searching for it:
- It is not FLUX.2 [max]. That was an image model from the previous FLUX.2 generation.
- It is not a "Max" plan. Several third-party sites sell subscription tiers called Max; those are price plans, not this model.
What FLUX 3 Max videos look like
fal published a set of FLUX 3 Max examples with the exact prompt behind each: one prompt, one pass, no editing. Three of them are below. They play muted as you scroll past; turn the sound on in the player to hear the audio FLUX 3 Max generated with each shot. Use this prompt opens the generator on this site with that prompt filled in. A 10-second 720p run like these costs 640 credits here.
Text to video · 10 seconds · 720p · sound on
One continuous ten-second Western shot set in the 1880s American frontier, with one piece of startling modern technology that feels out of place. A white driverless electric car with a roof-mounted spinning sensor pulls up silently in front of a frontier saloon with nobody at the wheel; a cowboy cautiously opens the rear door, peers in, and climbs in as the car glides away down the dusty street. Medium shot. Sunbaked western palette, hard sun, warm golden-sepia tones, soft dusty haze, rich 35mm film look with grain and halation, like a 1960s Western. Every person is an original fictional adult. No captions, logos or watermarks.
Use this prompt (opens in a new tab)Text to video · 10 seconds · 720p · sound on
One continuous ten-second Western shot. A pony-express rider at full gallop across open sagebrush country reaches a relay post; a waiting rider on a fresh horse rides alongside at speed and the leather mail pouch is passed from hand to hand without either horse slowing, then the second rider surges ahead. Side-on tracking shot at speed. Sunbaked western palette, hard sun, warm golden-sepia tones, soft dusty haze, rich 35mm film look with grain and halation, like a 1960s Western. Every person is an original fictional adult. No captions, logos or watermarks.
Use this prompt (opens in a new tab)Text to video · 10 seconds · 720p · sound on
One continuous ten-second Western shot set in the 1880s American frontier, with one piece of startling modern technology that feels out of place. Inside a frontier telegraph office, an old telegrapher taps his brass key and a glowing life-size hologram of a woman in a modern suit appears in the air in front of him, speaking; he slowly takes off his spectacles in amazement. Medium shot, warm light, the hologram glowing pale blue. Sunbaked western palette, hard sun, warm golden-sepia tones, soft dusty haze, rich 35mm film look with grain and halation, like a 1960s Western. Every person is an original fictional adult. No captions, logos or watermarks.
Use this prompt (opens in a new tab)Notice how the prompts are written: one continuous shot, the action in order, the camera, the look, and a closing line about what to avoid. That structure works for your own ideas too.
Is FLUX 3 Max different from FLUX 3 Video?
Yes. FLUX 3 Max is the post-trained, faster version of Black Forest Labs' FLUX 3 video model, which many sites call FLUX 3 Video. They share the same base, but they are sold differently and expose different options:
| FLUX 3 Max | FLUX 3 (Black Forest Labs' release) | |
|---|---|---|
| Who serves it | fal, and apps built on it | Black Forest Labs' API, and apps built on it |
| Modes | Text to video, image to video (start/end frame), keyframes | Text to video, image to video with keyframes, continuation, drafts |
| Length | 5 to 20 seconds | 5 to 20 seconds, or auto |
| Resolution | 720p or 1080p | hd, fhd, qhd, or uhd (up to 3840 × 2176 at 16:9) |
| Audio | On by default, can be switched off | On by default |
| 10-second clip time | 12.45 s (fal's test) | 126.44 s in fal's launch chart; 46.29 s in its later chart |
| Image-to-video Elo | 1,139 in fal's chart (preliminary); 1138 ±15 on Artificial Analysis today | 1,089 in fal's chart |
The FLUX 3 column comes from Black Forest Labs' API documentation; the timing and Elo figures come from fal's FLUX 3 Max page and TokenPost's report on fal's charts. On fal's numbers, post-training lifted quality by 50 Elo points and cut a 10-second clip from 126.44 to 12.45 seconds, about 10× faster.
How fast is FLUX 3 Max? Is it really 22× faster than Seedance 2.5?
The 12.45-second figure for a 10-second clip is consistent across fal's materials. The comparison is not:
- fal's launch chart put Seedance 2.5 at 277.80 seconds for the same 10-second, 16:9 clip, which is where "22× faster" comes from.
- A later fal chart kept FLUX 3 Max at 12.45 seconds but listed Seedance 2.5 at 69.09 seconds, which makes the gap about 5.5×. TokenPost reported the change on October 20, 2026; fal has not explained it.
- Black Forest Labs' page puts it as "10-second video in 12 seconds" and "22× faster than Seedance 2.5 at the same quality".
None of fal's charts say what hardware ran the test, whether queue time is included, or whether audio was on. Depending on which fal chart you trust, FLUX 3 Max is somewhere between about 5.5× and 22× faster than Seedance 2.5 on model time.
How fast is FLUX 3 Max in practice?
Two independent tests, both run through fal, measured the wait a user actually sees:
| Test | What they ran | FLUX 3 Max | Seedance 2.5 |
|---|---|---|---|
| Voyager, October 19, 2026 | Nine fixed prompts, 5 seconds each, 720p, audio on, fal's default prompt reasoning; waits include queue and polling | About 16 seconds per clip (one took 39); nine clips cost $1.91 | 3 to 26 minutes per clip (run in September); nine clips cost $18.49 |
| Promptowy, October 20, 2026 | One 10-second 720p clip, timed in three stages | Text to video: 9.3 s generating, 18.8 s in fal's pipeline, 20.5 s from submit to file. Image to video: 9.6 s, 21.2 s, 25.1 s | 197.5 s for a 480p clip |
So the model's own generation time does match fal's "about 12 seconds" claim, but the finished file takes roughly twice as long once prompt expansion, encoding, upload and the queue are counted. Both are small samples from single testers, not a benchmark.
On this site, a run also rewrites your prompt into a detailed shot description, then renders, uploads, and saves the clip to your history, so expect a similar end-to-end wait. We have not published timings of our own yet.
How good is FLUX 3 Max?
The only independent quality data is Artificial Analysis's image-to-video leaderboard, built from blind votes. Read on October 21, 2026, it shows FLUX 3 Max (720p) at 1138 ±15 Elo from 5,054 samples, next to MiniMax H3 (768p) at 1140 ±16 and Seedance 2.5 at 1138 ±17. The confidence intervals overlap, so the three are statistically tied. fal's launch material quotes 1,139 from the preliminary snapshot of the same arena. Nobody has published a hands-on review of its audio quality or its start-and-end-frame behaviour yet.
How much does FLUX 3 Max cost per second?
On fal, FLUX 3 Max is billed per second of output video, the same rate for text and image input:
| Resolution | List price per second | Launch price (75% off, until November 7, 2026) | 10-second clip at list price |
|---|---|---|---|
| 720p | $0.17 | $0.0425 | $1.70 |
| 1080p | $0.29 | $0.0725 | $2.90 |
fal's model documentation says the discount runs from October 15 to November 7. It does not give a time zone, so assume the list price applies from November 8. Read from fal's FLUX 3 Max model documentation on October 20 and 21, 2026.
On this site, FLUX 3 Max costs credits per second of video, and the price does not change with sound, aspect ratio, or a start or end frame:
| Length | 720p | 1080p |
|---|---|---|
| 5 seconds | 320 credits | 540 credits |
| 10 seconds | 640 credits | 1,080 credits |
| 20 seconds | 1,280 credits | 2,160 credits |
Our credit prices are set from fal's list price, not the launch discount, so they will not jump on November 8. A failed run refunds its credits automatically.
Can FLUX 3 Max make videos with sound, and how long can they be?
Yes. FLUX 3 Max generates audio together with the picture, and it is on by default; turn it off for a silent clip. A clip can be any whole number of seconds from 5 to 20. For longer pieces you would join several clips; FLUX 3 Max does not offer the continuation mode that Black Forest Labs' own FLUX 3 API has.
Can I turn my own image into a video with FLUX 3 Max?
Yes. Give it an image as the start frame and describe the motion. Add a second image as the end frame and the clip travels from one to the other. With a start frame, Auto keeps the image's own shape; without one, Auto renders 16:9. With three to ten images, FLUX 3 Max uses them as keyframes. On this site they are spread evenly through the clip in the order you add them: the first opens the clip and the last closes it. The price per second is the same however many frames you add.
Black Forest Labs' prompting advice for its FLUX 3 base model applies here too: keep the start and end images about the same subject and camera setup, because the further apart they are, the less predictable the motion between them.
Where can I use FLUX 3 Max online without the API or writing code?
These are the no-code routes we could confirm on October 21, 2026:
- fal's playground. fal's FLUX 3 Max page opens a playground in the browser. You need a fal account and pay per second of output, with no subscription.
- Krea. Krea has a FLUX 3 Max model page and generator. Its pricing lists all video models from the Pro plan ($35 a month) up; its separate plan named "Max" is a price tier, not this model.
- Layer. Layer's documentation lists FLUX.3 Max among the models it hosts.
- This site. The FLUX 3 Max generator runs it in your browser: sign in with Google, write a prompt or add a start and end frame, pick 720p or 1080p and a length, and pay in credits. The price shows on the prompt box before you run.
We found no free way to use FLUX 3 Max. fal bills every second, Krea's free plan does not list video models, and this site has no free tier.
We are an independent workbench, not affiliated with, endorsed by, or sponsored by Black Forest Labs or fal. FLUX is a trademark of Black Forest Labs, used here only to identify the models we write about and provide access to.
Timeline
- July 23, 2026: Black Forest Labs announces the FLUX 3 family, starting with video.
- October 15, 2026: fal's launch discount on FLUX 3 Max starts (75% off).
- October 16, 2026: FLUX 3 Max text-to-video, image-to-video and keyframes endpoints appear on fal.
- October 19, 2026: launch coverage of FLUX 3 Max, for example "Fal Debuts FLUX 3 Max Video Model Post-Trained With Black Forest Labs" on Unite.AI.
- October 19, 2026: Voyager publishes the first independent side-by-side test against Seedance 2.5.
- October 20, 2026: TokenPost reports that fal's two speed charts disagree on Seedance 2.5 (277.80 vs 69.09 seconds); Promptowy publishes an end-to-end timing.
- November 7, 2026: the launch discount ends, per fal.
FAQs about FLUX 3 Max
What is FLUX 3 Max?
A video model made by fal Research in collaboration with Black Forest Labs, post-trained from Black Forest Labs' FLUX 3 video model. It makes 5 to 20 second clips with sound at 720p or 1080p from text, or from a start frame with an optional end frame.
Is FLUX 3 Max real, or just a plan name?
It is a real model, live on fal since October 16, 2026, and featured on Black Forest Labs' FLUX 3 page. Some unrelated sites also sell a "Max" subscription plan; that is a different thing.
Is FLUX 3 Max different from FLUX 3 Video?
Yes. FLUX 3 Video is Black Forest Labs' own FLUX 3 model with drafts, keyframes, continuation and up to uhd resolution. FLUX 3 Max is fal Research's post-trained version, focused on text to video and image to video at 720p or 1080p, and about 10× faster on fal's numbers.
Is FLUX 3 Max really 22× faster than Seedance 2.5?
Only on fal's launch chart: 12.45 seconds against 277.80 for a 10-second clip. A later fal chart lists Seedance 2.5 at 69.09 seconds, about 5.5×. In independent tests, FLUX 3 Max returned 5-second clips in about 16 seconds including the queue, while Seedance 2.5 took 3 to 26 minutes, and a 10-second clip took 20.5 seconds from submit to file.
How much does FLUX 3 Max cost per second?
$0.17 per second at 720p and $0.29 at 1080p list on fal, or $0.0425 and $0.0725 during the 75% launch discount that ends November 7, 2026. On this site it is 64 credits per second at 720p and 108 at 1080p.
Can FLUX 3 Max make videos with sound?
Yes. Audio is generated with the picture and is on by default. You can switch it off.
How long can FLUX 3 Max videos be?
5 to 20 seconds per clip, in whole seconds.
Can FLUX 3 Max turn an image into a video with a start and end frame?
Yes. One image sets the start frame, a second image sets the end frame, and three to ten images become keyframes spread evenly through the clip. The price per second does not change with the number of frames.
Is FLUX 3 Max an image model?
No. For still images, use FLUX 3 Image or FLUX 3 Image Klein on this site.
Can I use FLUX 3 Max for free?
We found no free route as of October 21, 2026. fal charges per second, and this site uses paid credits with automatic refunds for failed runs.
Sources: fal, fal.ai/flux-3-max and the FLUX 3 Max model documentation and schemas (blackforestlabs/flux-3-max/text-to-video, image-to-video, keyframes-to-video); Black Forest Labs, bfl.ai/models/flux-3-video and the FLUX 3 video API documentation (docs.bfl.ai); Artificial Analysis image-to-video leaderboard; TokenPost, October 20, 2026; Unite.AI, October 19, 2026; Voyager, "FLUX 3 Max vs Seedance 2.5", verified October 19, 2026; Promptowy, "FLUX 3 Max: 10 s wideo w 12 s? Mój pomiar", October 20, 2026; Krea, krea.ai/models/flux-3-max; Layer documentation. Read October 20–21, 2026.