FLUX 3 Max
FLUX 3 Max is the newest post-trained video model from fal Research, built with Black Forest Labs on FLUX 3. Tied for #1 in image-to-video quality, and a 10-second video in about 12 seconds.
Your prompt opens in the FLUX 3 Max playground, ready to generate.
Frontier quality at a fraction of the latency
Tied for #1 in image-to-video quality
FLUX 3 Max scores 1,139 Elo in Artificial Analysis's preliminary image-to-video arena, tied for #1 in blind human preference votes and ahead of Gemini Omni Flash 1.1, Wan 3.0 and H3 Max.
Seedance 2.5 quality, 22x faster
For a 10-second clip, FLUX 3 Max matches Seedance 2.5 on quality, with both at 1,139 Elo, and generates it about 22 times faster: 12.45 seconds against 277.80.
What you can make with FLUX 3 Max
Speed so fast it feels out of place
A 10-second clip comes back in about 12 seconds, close to real time, at a quality other models take minutes to reach. That is about 22x faster than Seedance 2.5, so you can try twenty takes of a shot in the time it returns one.
Unreal speed and quality
FLUX 3 Max is tied for #1 in image-to-video quality on Artificial Analysis's preliminary arena, and it is one of the fastest video models on fal. You no longer need a fast model for drafts and a slow one for finals: the take you iterate on is the take you ship.
Switch between the best models, post-trained by fal
H3 Max and FLUX 3 Max are both post-trained by fal Research and served on the same API, so you can move a shot from one to the other by changing a single model ID. Use each where it is strongest, with no new integration.
Faces that hold up in close-up
An extreme low angle on the station keeper as she tips her hat brim up a finger's width. Skin, sweat-stained felt and the pale sky behind her stay natural at a distance where most models start to slip.
Motion that stays true to life. Made for detail.
Fast-moving subjects keep their shape from the first frame to the last: horses hold their stride, wheels keep their spokes, painted wood keeps its grain and dust trails the way it should. Text to video and image to video share the same post-trained model, so that fidelity carries across both at 720p.
Let your ideas take flight
Describe the shot no one has filmed: a rocket booster settling onto a frontier ranch at dawn, dust rolling past a cowboy at the fence. FLUX 3 Max follows the whole idea, from the scale of the landing to the light on the canyon walls, so the strangest pitch comes back looking like it was shot on location.
How fast is FLUX 3 Max?
FLUX 3 Max generates a 10-second video in 12.45 seconds and scores 1,139 Elo on Artificial Analysis's image-to-video arena, tied for #1 in the preliminary snapshot. Seedance 2.5 scores the same and takes 277.80 seconds.
Better and faster than FLUX 3
Post-training lifted FLUX 3 by 50 Elo points, from 1,089 to 1,139, while inference work cut a 10-second video from 126.44 seconds to 12.45: about 10x faster.
Pushing the speed and quality frontier
Plot quality against latency and FLUX 3 Max sits top left, next to H3 Max: no other model in the comparison is both higher rated and faster.
| Model | Image-to-video Elo | Time for a 10s video |
|---|---|---|
| FLUX 3 Max | 1,139 | 12.45s |
| Gemini Omni Flash 1.1 | 1,134 | 47.37s |
| MiniMax H3 (768p) | 1,141 | 83.26s |
| FLUX 3 | 1,089 | 126.44s |
| Wan 3.0 | 1,127 | 180.16s |
| Seedance 2.5 | 1,139 | 277.80s |
Elo: Artificial Analysis AA-Video-I2V v1.1, preliminary. FLUX 3 Max is tied for #1: its confidence interval overlaps the leaders, with MiniMax H3 listed first on that board at 1,141. Sorted fastest first. Generation time: average for a 10-second, 16:9 video at 720p/768p, in seconds.
FLUX 3 Max is 75% off until November 7
Billed per second of output video, with no minimums and no subscription. Text to video and image to video cost the same.
| Resolution | FLUX 3 Max |
|---|---|
| 720p | $0.0425 / secondList $0.17, this rate ends November 7 |
See what FLUX 3 Max can create
One prompt each, one pass, no editing. Open any prompt in the playground to run it yourself.
The mail at full gallop
"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."
The brim lift
"Create one continuous six-second Western shot of the brim lift: an extreme low-angle close-up of the station keeper against the pale sky, her hat brim cropping the top of frame, as she tips it up at the first sight of the rider's dust. Calm, luminous, natural. A lone 1870s horse-relay station on a high bleached mesa top at noon: cracked bone-tan clay, grey-green sage, silvered juniper snags, far pale ochre mesas under a faded sky-blue sky. An adobe station house with a sagging grey-timber porch, a split-rail corral, a tarred plank trough and a crooked cedar telegraph pole line. Everyone wears patched, sun-rotted cotton, buckskin, canvas and wool, sweat-stained felt hats, dust in every seam; brass is tarnished, iron hand-forged. Dry tumbleweeds bounce across the open ground on a hot wind. Nothing is new: every board, rope, saddle and cloth is decades-worn and mended. Every person is an original fictional adult. At the porch rail, filling the lower frame against the sky, stands Honora Coyle, a fictional 44-year-old widow who keeps the relay station: strikingly beautiful in a natural classic-film-star way, fair skin with a soft natural tone and only a light dew of sweat, strong cheekbones, calm grey-green eyes with deep squint lines, tall and straight-backed, dark auburn braid tied with a frayed faded-blue ribbon. She wears a sun-bleached, patched blue calico work dress with rolled sleeves, a stained tan canvas apron, a sweat-stained bone felt hat and cracked brown boots. A sun-greyed porch post stands soft in the near right foreground. Camera: a low dolly on a short track below the porch rail, looking steeply up at her face against the sky. 50mm lens impression, no wide-angle distortion. One very slow forward creep of forty centimetres. Because the camera inches closer from below, the post slides out of the right edge and her brim comes to crop the top of frame as a hard horizontal line. Not a zoom, not a tilt, not handheld. Timing: seconds zero to two, Honora stands still against the faded sky, the brim's shadow soft across her brow, skin natural and matte with a faint dew at the temple. Seconds two to four, her eyes shift far off camera-left. Seconds four to five, with two fingers she tips the brim up a finger's width and her eyes narrow. Final second, she holds still, brim edge level across the top of frame. End on her eyeline. A frame from a 1950s-60s Western feature film shot on location on 35mm colour negative and seen as an aged dye-transfer release print: rich but time-worn colour, deep faded-blue skies, sunburnt earth reds and ochres, warm leathery skin, dense shadows, coarse organic film grain, halation around bright edges, the soft focus fall-off and light vignetting of vintage lenses, faint print dust. Big open vistas and low-angle close-ups against the sky; tripod and dolly only. Hard noon sun overhead, camera-left. Sweat and dust on real sun-leathered skin. No TV-screen frame, rounded corners, scanlines, B&W, HDR or digital crispness. 16:9, 24 fps. No captions, logos or watermarks. Avoid: orange or shiny skin, heavy sweat, harsh glare, smiling, looking into the lens, a second person, rounded frame corners or black borders."
A driverless car at the saloon
"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."
A robot dog working the herd
"One continuous ten-second Western shot set in the 1880s American frontier, with one piece of startling modern technology that feels out of place. On a frontier cattle ranch, a four-legged robot dog trots alongside a mounted cowboy, nimbly herding a few stray longhorn cattle back toward the corral through the sagebrush; the cowboy tips his hat to it. Wide tracking 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."
A call from the future
"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."
Muscle from the future
"One continuous ten-second Western shot set in the 1880s American frontier, with one piece of startling modern technology that feels out of place. At the mouth of a frontier silver mine, a miner wearing a sleek powered exoskeleton frame over his work clothes lifts a fully loaded ore cart clean off the rails with ease and sets it down, while the other dusty miners with pickaxes stop and stare with their mouths open. Medium-wide 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."
Post-trained by fal Research, with Black Forest Labs
FLUX 3 Max starts from Black Forest Labs' FLUX 3. fal Research worked on three layers at once, because a gain at one layer opens room at the next, and is continuing to post-train the model.
Post-training
Inference optimization
Systems engineering
Check out other post-trained models from fal
FLUX 3 Max comes from the same fal Research team that built H3 Max.Explore H3 Max
H3 Max
fal's post-trained MiniMax H3: text and image to video with native audio, tuned for prompt adherence and aesthetics. A 5-second 768p clip renders in under 3 seconds. $0.08 per second at 768p, with H3 Max Turbo at half that, pay-per-use with no subscription. See H3 Max.
Build with the FLUX 3 Max API
The client API handles the request submit protocol. It will handle the request status updates and return the result when the request is completed.
import { fal } from "@fal-ai/client";
const result = await fal.subscribe("blackforestlabs/flux-3-max/text-to-video", {
input: {
prompt:
"A single-file line of longhorns swims a cold jade river at dawn, only heads and wide pale horns above the water, riders on the far bank. Wide, still, enormous.",
},
logs: true,
});
console.log(result.data.video.url);
Common questions about FLUX 3 Max
What is FLUX 3 Max?
FLUX 3 Max is a post-trained video model from fal Research, built in partnership with Black Forest Labs on FLUX 3. It is tied for #1 in image-to-video quality on Artificial Analysis's preliminary arena and generates a 10-second video in about 12 seconds, at 720p. fal Research is continuing to post-train it.
What is a post-trained model?
A post-trained model starts from an existing base model and keeps training it on new data and objectives, rather than training from scratch. The base model supplies its general knowledge of motion, light and composition; post-training sharpens specific behaviors on top of it, such as following prompts more closely or producing more polished video. fal Research pairs post-training with inference optimization and systems engineering, so a model can get better and faster at the same time.
Who made FLUX 3 Max?
fal Research made FLUX 3 Max in collaboration with Black Forest Labs, who built the base model, FLUX 3. Black Forest Labs brought the foundation model; fal Research brought post-training, inference optimization and systems engineering. It comes from the same team that built H3 Max.
How much does FLUX 3 Max cost?
FLUX 3 Max is 75% off until November 7, so it runs at $0.0425 per second at 720p. List price is $0.17 per second at 720p. Billing is per second of output video, with no minimums and no subscription. Current pricing is listed on the fal.ai pricing page.
Which FLUX 3 Max endpoints are available?
Two at launch: text to video and image to video. Information updated as of October 18, 2026.
How is FLUX 3 Max different from FLUX 3?
FLUX 3 is Black Forest Labs' own release, with draft, keyframe and extend endpoints. FLUX 3 Max is the post-trained version, focused on text to video and image to video. In Artificial Analysis's preliminary image-to-video arena it scores 50 Elo points higher than FLUX 3 (1,139 against 1,089), and it generates a 10-second video about 10 times faster (12.45 seconds against 126.44).
How fast is FLUX 3 Max?
FLUX 3 Max generates a 10-second video in 12.45 seconds, against 277.80 seconds for Seedance 2.5: about 22 times faster at the same 1,139 Elo in Artificial Analysis's preliminary image-to-video arena. The only faster model in that comparison is H3 Max, fal Research's other post-trained model, at 7.55 seconds.
How do I get started with the API?
Install the fal.ai SDK (Python or JavaScript), grab an API key from your dashboard, and make your first request in a few lines of code. Check the API documentation for every available parameter.
Can I use FLUX 3 Max for commercial projects?
Yes. Content generated through the fal.ai API can be used in commercial projects. Check fal.ai's terms of service for full details on usage rights and licensing.
Get in touch about FLUX 3 Max
Want to learn more about integrating FLUX 3 Max into your workflow? Leave your details and our team will reach out.