FLUX 3 Max Review: fal's 12-Second AI Video Model, 75% Off Until Nov 7

On October 19, 2026, fal launched FLUX 3 Max, a video model built on Black Forest Labs' FLUX 3 and post-trained in-house by fal Research. The pitch is simple: a 10-second 720p clip in about 12 seconds, at 75% off launch pricing until November 7. I have spent the last two days reading the announcement, the product page, and every piece of independent coverage I could find, because a launch-day discount with a two-week window is exactly when you want the fine print first.

Quick answer: FLUX 3 Max generates 10 seconds of 720p AI video in roughly 12 seconds, according to fal's own benchmarks. Launch pricing is $0.0425 per second of output through November 7, 2026, then $0.17 per second. Independent preference rankings put it near the top of image-to-video quality but statistically tied with rivals — so the real story is speed plus a discount, not a proven quality crown.

What FLUX 3 Max actually is

FLUX 3 Max is not a brand-new foundation model. It starts from Black Forest Labs' FLUX 3 video model, announced in July 2026, and fal applied its own post-training: additional training with new data and objectives, plus reinforcement learning aimed at motion quality, visual fidelity, and prompt-following. fal's own description of the split is explicit: "Black Forest Labs brought the foundation model; fal Research brought post-training, inference optimization and systems engineering."

It is served exclusively through fal's API platform — there is no open-weight release, and the cited launch materials establish no rights to download, modify, or operate the model independently. This matters if you were hoping to self-host: you cannot.

Three generation modes are advertised: text-to-video, image-to-video, and keyframes-to-video, where you supply a first and last frame and the model renders the motion between them. fal's product page lists text-to-video and image-to-video as the two endpoints live at launch, both at 720p, with a 1080p upscale option. Clips run 5 to 20 seconds, native audio is on by default, and the aspect-ratio list is unusually complete for a launch-week model: 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16. A "reasoning-level" setting automatically expands short prompts into more detailed ones, a small but practical touch for people iterating quickly.

Speed: what "22× faster" really means

The headline claim is the whole launch. fal says FLUX 3 Max generates a 10-second clip in 12.45 seconds, compared with 277.80 seconds for Seedance 2.5 — about 22 times faster — and 126.44 seconds for the original FLUX 3, roughly 11.5 times faster. Against the original model, fal says post-training "lifted FLUX 3 by 50 Elo points, from 1,089 to 1,139" while cutting generation time to a tenth.

Two caveats, both important. First, every timing figure here comes from fal — there are no independent speed tests yet. Second, fal's own numbers do not fully agree with each other: the launch announcement reports 10.95 seconds for a 10-second clip, while the product page lists a 12.45-second average. The 12.45-second figure also appears in VentureBeat's reporting of fal's materials. Treat the speed story as "roughly 12 seconds, per fal's testing" and expect your own results to vary with settings and operating conditions, as Superpower Daily's coverage notes.

The product page's comparison table puts the claim in context against named competitors, all timed by fal for a 10-second 16:9 video: Gemini Omni Flash 1.1 at 47.37 seconds, MiniMax H3 (768p) at 83.26 seconds, Wan 3.0 at 180.16 seconds, Seedance 2.5 at 277.80 seconds. fal's own benchmark, fal's own conditions — noted.

Quality evidence: the Elo picture is a tie

For quality, fal leans on the Artificial Analysis Video Arena, an independent blind-preference leaderboard. fal reports FLUX 3 Max reached a 1,139 Elo rating across 4,868 votes, calling it a "joint #1" in image-to-video quality while noting that MiniMax H3 is listed first and that rankings may shift as more votes come in.

Look at the actual board and the claim softens. The leaderboard listed FLUX 3 Max at 720p with 1,138 Elo (±15) across 5,054 samples, ranked second behind MiniMax H3 at 768p with 1,140 Elo. Seedance 2.5 sat at 1,138, the original FLUX 3 at 1,086. The confidence intervals of the leading models overlap — statistically, they are tied. As Superpower Daily put it, the evidence does not establish a clear winner. And this evaluation covers image-to-video with audio only; nothing here independently evaluates the text-to-video mode.

So: quality is genuinely top-tier, but "best" is not proven — it is fal's description of a preliminary, tied result. That distinction is the difference between marketing copy and a review.

Pricing, with the fine print

This is the section most launch coverage glosses over, so let us be exact. fal lists a promotional price of $0.0425 per second of 720p output, billed per second with no minimums and no subscription required. It is 75% off the $0.17-per-second list price, and it ends November 7, 2026. The 1080p upscale costs $0.0725 per second on promotion and $0.29 after. fal's release describes the discount window as the first three weeks; coverage notes the two companies describe it as two to three weeks, so mark your calendar rather than trusting the exact count.

In practical terms: a 10-second 720p clip costs about $0.43 during the promotion and $1.70 after. A team that generates 10 versions of a 10-second shot — completely normal when iterating on a scene — spends $4.25 at promo pricing and $17 at the regular rate. Staff time and editing are not in that math, and they never are.

Two more details that matter for buyers. First, fal says content generated through its API may be used in commercial projects under fal's terms — that permission should not be confused with any right to the model itself. Second, this mirrors fal's established playbook: fal's H3 Max is a similarly post-trained version of MiniMax's H3, served on the same platform. This is a product strategy now, not a one-off.

How it stacks up

The table makes the honest point: FLUX 3 Max is the speed outlier in this group by fal's own measurements, but it is not the price outlier — even at launch pricing, a MiniMax option undercuts it per second on fal's own platform, and the promo price is the only thing that separates it from the regular $0.17-per-second rate it shares with the original FLUX 3.

Pros and cons

Pros

• Roughly 12 seconds per 10-second clip, per fal's benchmarks — an order of magnitude faster than the named competitors it was measured against.
• Independent preference scores place it in the top tier for image-to-video, statistically tied with Seedance 2.5 and MiniMax H3.
• Launch pricing of $0.0425/sec through November 7, billed per second with no minimums or subscription — genuinely cheap to experiment with.
• Broad aspect-ratio support (21:9 to 9:16), 5–20 second clips, native audio, and keyframes mode for storyboard-style control.
• Commercial use of API-generated content is permitted under fal's terms.

Cons

• All speed figures come from fal's own testing; no independent benchmarks exist yet, and fal's announcement and product page disagree slightly (10.95s vs 12.45s).
• The "best quality" claim rests on a tied, preliminary Elo result whose confidence intervals overlap rivals — and it covers image-to-video only.
• The promo price expires November 7, 2026; the list price is 4× higher at $0.17/sec, matching the original FLUX 3.
• fal-API-only — no open weights, no self-hosting, and no confirmed end date flexibility on the discount.
• Faster generation can mean more versions, not lower spend: iterate freely and the 75% discount still bills per second.

Who should try it this week

If you storyboard, iterate on product shots, or build pipelines that call a video model repeatedly, the combination of a ~12-second turnaround and a two-week discount is a genuine reason to run a controlled test now. The workflow question is concrete: can your team reach an acceptable shot faster, at a predictable cost, while keeping the details the brief requires? fal's own suggestion — and the sensible one — is to test that on your own briefs rather than on vendor benchmarks.

If you need text-to-video specifically, or you plan to build long-term infrastructure on one model, waiting two weeks has a real cost: the price quadruples on November 8, and independent testing may arrive in the meantime. The discount rewards early movers; it does not obligate you to be one.

For the broader landscape of what happened when a major video model disappeared entirely this year — and which tools teams switched to — see our guide to Sora alternatives after the Sora 2 shutdown.

FAQ

How much does FLUX 3 Max cost?

$0.0425 per second of 720p output through November 7, 2026 (75% off the $0.17 list price), billed per second with no minimums and no subscription. A 10-second clip costs about $0.43 on promotion and $1.70 after. The 1080p upscale is $0.0725/sec promo, $0.29/sec list.

How fast is FLUX 3 Max?

fal reports 12.45 seconds to generate a 10-second 720p clip, about 22× faster than Seedance 2.5's 277.80 seconds in its own comparison. The launch announcement separately reported 10.95 seconds. Both figures are fal's own; independent speed tests have not been published.

Can I use FLUX 3 Max videos commercially?

fal states that content generated through its API may be used in commercial projects under fal's terms. That is a usage permission on the outputs, not a license to the model itself — there is no open-weight release.

How do I access FLUX 3 Max?

Through fal's API at fal.ai/models/fal-ai/flux-3-max, with text-to-video and image-to-video endpoints live at launch. There is no consumer app and no self-hosted option.

Is FLUX 3 Max better than Seedance 2.5?

On fal's own measurements it is roughly 22× faster at a tied preference score (both ~1,138–1,139 Elo). But the quality ranking is preliminary and statistically tied across the leading models, and no independent speed testing exists — "faster per fal, quality a draw so far" is the honest summary.

Key takeaways

• FLUX 3 Max, launched October 19, 2026, renders a 10-second 720p clip in about 12 seconds per fal's own benchmarks — roughly 22× faster than Seedance 2.5 on the same test.
• Launch pricing of $0.0425 per second through November 7 makes a 10-second clip about $0.43; the price returns to $0.17 per second afterward, matching the original FLUX 3.
• Independent blind-preference scores place it in the top tier for image-to-video but statistically tied with Seedance 2.5 and MiniMax H3 — the quality crown is unproven.
• It is fal-API-only with no open weights, but API-generated content may be used commercially under fal's terms.
• The strongest case is a two-week workflow test: whether faster generations actually shorten your revision cycles at a cost you can predict.

Published October 21, 2026. Prices and benchmarks verified against fal's product page and independent reporting on October 20–21; all speed and Elo figures are vendor-reported or preliminary unless stated otherwise. The promo price ends November 7 — check fal's page before you spend.