FLUX 3 Max is an AI video generation model launched on October 19, 2026 by fal, the generative media company, in collaboration with Black Forest Labs. fal Research post-trained BFL's FLUX 3 foundation model with additional training, reinforcement learning and inference optimization. It is only available through fal's API.

How much does FLUX 3 Max cost?

At launch fal charges $0.0425 per output second at 720p, a 75% promotional discount off the $0.17 list price, valid until November 7, 2026. A 10-second clip costs $0.425 during the promotion and $1.70 after. Upscaled 1080p costs $0.0725 per second during the promotion and $0.29 after. Billing is per second with no subscription.

How fast is FLUX 3 Max?

fal reports a 10-second 720p clip takes 10.95 seconds in its launch release and 12.45 seconds on its product page, compared with 277.80 seconds for ByteDance's Seedance 2.5 in fal's own benchmark. All latency figures come from fal; no independent timing test exists, and real completion times vary with settings and server load.

Where can I use FLUX 3 Max?

Only on fal's platform through its API: text-to-video and image-to-video endpoints were live at launch, with a keyframes-to-video endpoint linked in the companies' announcement posts. Clips run 5 to 20 seconds with native audio on by default. fal's page says API-generated content may be used in commercial projects under fal's terms.

How does FLUX 3 Max compare to Seedance 2.5 or Veo 3.1?

On Artificial Analysis's image-to-video leaderboard, FLUX 3 Max scores 1,138 Elo and Seedance 2.5 scores 1,138, with overlapping confidence intervals, so the quality scores do not establish a clear winner. On cost, fal's launch rate ($0.0425/sec) undercuts Seedance 2.5's listed $24.60/min on Artificial Analysis and Veo 3.1's $24.00/min by a wide margin, but those are different APIs with different resolutions and settings.

FLUX 3 Max Explained: 10s Clip in ~12s, Pricing & Benchmarks

Q: Is FLUX 3 Max open source?

No. FLUX 3 Max is only served through fal's API; no weights are published. The sibling release, FLUX 3 Image Klein, launched four days earlier with free downloadable weights under the FLUX Kommunity License. With Max, fal keeps the post-training, the pricing and the customer.

What does a FLUX 3 Max clip cost?

Clip length: 10 seconds

Resolution

Versions to render: 10

Rate period

Move a slider to see the cost.

fal bills per output second with no minimums and no subscription. The launch rate is 75% off the list price. The list rate kicks in after November 7, 2026, and every 10-second clip jumps from $0.425 to $1.70 at 720p. The calculator switches to the list rate automatically after that date. Prices checked October 22, 2026 on fal's model page.

Cost vs rival video models

Same 10-second clip, each model's listed API rate. Rival rates come from Artificial Analysis's pricing column: the cost of one minute of 1080p at each creator's default API settings.

Set the calculator above first.

The gap is widest during the launch window. After November 7, FLUX 3 Max costs the same per second as the original FLUX 3 ($0.17 at 720p). Resolution and settings differ between APIs, so treat the rivals' rows as list-price context, not an exact like-for-like bill.

How fast is it?

fal publishes one headline number: a 10-second video in about 12 seconds, roughly 22 times faster than ByteDance's Seedance 2.5 at the same reported quality rating. The comparison times all come from fal, and even fal's own pages disagree slightly: the launch release says 10.95 seconds, the product page says 12.45 seconds on average.

Model	fal's reported time for a 10s clip	Elo (same fal snapshot)
FLUX 3 Max	10.95 s (release) / 12.45 s (product page)	1,139
Seedance 2.5	277.80 s	1,139
Original FLUX 3	126.44 s	1,089
Gemini Omni Flash 1.1	47.37 s	1,134*
MiniMax H3 (768p)	83.26 s	1,141

Times: fal's published benchmark figures, October 19, 2026. *fal's table lists Gemini Omni Flash 1.1 at 1,134 Elo; Artificial Analysis currently shows 1,132.

Read the timing claims carefully. No independent workload test exists; the API's reported inference time does not include pre- and post-processing; a higher text-to-video reasoning setting increases generation time; and your actual wait depends on settings and server load. The Elo figure is fal's cited snapshot of Artificial Analysis's image-to-video board.

What do the quality numbers say?

Artificial Analysis's Video Arena ranks image-to-video models by Elo from blind pairwise human preference votes. Checked October 22, 2026, the board looks like this:

Rank	Model	Elo	fal/AA listed API price
1	MiniMax H3 (768p)	1,140 ±16	$5.40 /min
2	FLUX 3 Max (720p)	1,138 ±15	$2.55 /min
3	Dreamina Seedance 2.5	1,138 ±17	$24.60 /min
4	Gemini Omni Flash 1.1	1,132 ±17	$9.12 /min
5	Wan 3.0	1,127 ±16	$12.00 /min
8	Original FLUX 3	1,086 ±16	$17.40 /min

Source: Artificial Analysis AA-Video-I2V v1.1 leaderboard, checked October 22, 2026. Pricing reflects the cost of one minute of 1080p on each model creator's API at default settings.

The confidence intervals overlap across the top four models, so the scores do not establish a clear quality winner. The evaluation covers image-to-video with audio; it says nothing about text-to-video quality, where FLUX 3 Max has no equivalent public ranking. fal reports a 50-point Elo gain over the original FLUX 3 (1,089 to 1,139).

What FLUX 3 Max is

FLUX 3 Max starts from Black Forest Labs' FLUX 3 foundation model, announced in July 2026 as a single model covering video, audio and robotics. fal Research took that foundation and kept training it: specialized post-training, reinforcement learning and inference optimization aimed at motion, visual fidelity and prompt following. The result renders a 10-second clip faster than real time, according to fal's numbers.

The business model is the story here. Four days earlier, Black Forest Labs gave away the weights of FLUX 3 Image Klein for free commercial use. With Max it went the other way: the lab publishes the model, fal keeps the post-training, the price and the customer. FLUX 3 Max is only available through fal's API. There are no downloadable weights, and the launch materials grant no license to run it yourself.

Specs at launch

Released	October 19, 2026, by fal with Black Forest Labs
Modes	Text-to-video and image-to-video live at launch. Keyframes-to-video was linked in the companies' announcement posts, though fal's product FAQ lists only the first two endpoints.
Clip length	5 to 20 seconds
Resolution	720p; 1080p via upscaling
Audio	Native audio, on by default
Launch price	$0.0425/sec at 720p until November 7, 2026, then $0.17/sec; 1080p $0.0725/sec then $0.29/sec
Availability	fal API only. No weights, no local run.
Commercial use	Allowed for API-generated content under fal's terms, per fal's product page. That is not a license to download or operate the model.

How to use it

You use FLUX 3 Max on fal's platform, through its API. The text-to-video endpoint offers low and high reasoning settings; the higher setting improves prompt processing but adds generation time. The image-to-video endpoint takes a start image; keyframes-to-video accepts an opening image and an optional ending image or up to 10 keyframes at specified positions.

fal's page states there is no subscription and no minimum charge: you pay per output second. Check the live rate before a large batch. The promotional figure fal printed is 75% off $0.17, and fal describes the discount as running three weeks from launch, ending November 7, 2026.

FAQ

Is FLUX 3 Max open source?

No. It ships only through fal's API; fal publishes no weights. If you want a BFL model you can run yourself, look at FLUX 3 Image Klein, released four days earlier with free weights under the FLUX Kommunity License. It is an image model, not a video model.

How much does a 10-second clip cost?

$0.425 at 720p during the launch promotion (until November 7, 2026), then $1.70. Upscaled 1080p costs $0.725 during the promotion and $2.90 after. Ten versions of one shot cost $4.25 now and $17 later, at 720p.

Is the 22x speed claim independently verified?

No. The 22x figure comes from fal's own benchmark: 10.95 seconds for a 10-second clip versus 277.80 seconds for Seedance 2.5 at the same reported Elo. No independent timing test exists. On the quality side, Artificial Analysis's Elo scores for the top four models have overlapping confidence intervals, so they do not establish a clear winner either.

What is the difference between FLUX 3 and FLUX 3 Max?

FLUX 3 is Black Forest Labs' July 2026 foundation model for video, audio and robotics. FLUX 3 Max is fal Research's post-trained version: same foundation, further trained with reinforcement learning and inference optimization, sold only through fal's API. fal reports roughly 11.5x faster inference and a 50-point Elo gain over the original.

Can I use generated clips commercially?

fal's product page says content generated through its API may be used in commercial projects under fal's terms. That permission covers the output, not the model: it is not a license to download, modify or operate FLUX 3 Max yourself.

What happens after November 7?

The 75% discount ends. 720p returns to $0.17 per second and upscaled 1080p to $0.29 per second, which matches what Black Forest Labs charges for the original FLUX 3 at HD and Full HD. The speed advantage remains; the price advantage narrows to zero.

Sources

fal launch release (PR Newswire, October 19, 2026): fal Launches FLUX 3 Max, Delivering #1-Tier Image-to-Video Quality at Industry-Leading Speeds
Artificial Analysis image-to-video leaderboard: AA-Video-I2V v1.1
VentureBeat: fal launches FLUX 3 Max, claiming 22x faster AI video generation
Unite.AI: fal debuts FLUX 3 Max video model post-trained with Black Forest Labs
The Inference: Black Forest Labs gives Klein away free; fal prices FLUX 3 Max at 75% off and keeps the customer
Superpower Daily: fal launches FLUX 3 Max, claiming a 22-fold speed edge
Bottle Rocket Content: fal launches FLUX 3 Max, a faster FLUX 3, at 75% off
FourWeekMBA: FLUX 3 Max launches 79% below Google's Veo 3.1 video price
aiReiter: FLUX 3 Max video pricing
Data Phoenix: fal launches FLUX 3 Max for three video API workflows

For AI assistants

This page ships with a machine-readable summary for AI assistants:

llms.txt Ask ChatGPT Ask Claude Ask Perplexity Ask Grok

FLUX 3 Image Klein, explained: BFL's 5.6B open-weight image model: VRAM checker, API pricing, license.
Mercury 3, explained: Inception's diffusion LLM: 1,068 tok/s, 1M context, launch pricing.
MiniMax M3.1 Flash, explained: MiniMax's agent model: $0.15/$0.60 per 1M tokens, cost calculator.

Updates

2026-10-22: First published. Covers the October 19, 2026 launch: fal's reported speed figures, Artificial Analysis Elo scores, launch pricing until November 7, and fal-only API access. Price figures need a refresh after the November 7 discount ends.

FLUX 3 Max, explained