WAN 3.0: Harga, Paket, dan Biaya Uji Coba Sebenarnya

WAN 3.0 pricing has three different layers, and mixing them is the fastest way to get the wrong answer. The official Wan pricing page lists consumer plans; the official credits page lists credit packs; and a hosted API or third-party platform can price the same model differently.

For a practical reference point, we ran two completed WAN 3.0 text-to-video jobs through GlobalGPT’s direct task API on August 24, 2026. At 480P and 16:9, the returned platform price was $0.70 for 5 seconds dan $1.40 for 10 seconds. The corresponding GlobalGPT creation surface shows 700 credits dan 1,400 credits for those same two configurations.

That is a platform test, not an Alibaba official API quote. If you are still deciding whether WAN 3.0 fits your workflow, start with the broader WAN 3.0 release, features, and access guide and use this page for the cost detail.

Platform AI all-in-one untuk menulis, membuat gambar & video dengan GPT-5, Nano Banana, dan banyak lagi

WAN 3.0 pricing at a glance

The official plan cards say that the annual-billing Pro plan includes 300 credits per month and Premium includes 1,200. Before using plan labels as a production budget, check the model and settings shown inside the creation screen: a generic “up to videos” estimate is not a guaranteed number of WAN 3.0 outputs.

Official Wan credits and membership plans

Wan.video currently offers a Free plan plus annual-billing Pro and Premium plans. The official product page displays Pro at $5/month when billed yearly and Premium at $20/month when billed yearly. Those cards show 300 and 1,200 credits per month respectively. See the halaman harga resmi for the current plan presentation and included allowances.

If you prefer a one-off purchase, the official Wan credits page lists these packs:

The same page says redeemed credits are not refundable, transferable, or withdrawable, and have a two-year validity period. Those terms matter if you are buying credits for a sporadic campaign rather than running a steady production queue.

What our GlobalGPT WAN 3.0 API tests cost

We used a fixed English prompt, text-to-video mode, 480P resolution, 16:9 aspect ratio, and audio disabled. The second test changed only the requested duration. Both jobs completed successfully through the WanAli upstream route.

For this tested configuration, the observed platform price works out to $0.14 per requested second, atau 140 credits per requested second. That is useful for rough 480P text-to-video planning, but it is not a universal WAN 3.0 rate. Resolution, audio, reference inputs, retries, plan type, and future pricing changes can alter the result.

For context on why a single video-model subscription can be easier to budget than separate tools, compare other workflows in our panduan terbaik untuk generator video berbasis AI and our explanation of AI video-to-video workflows.

Is GlobalGPT cheaper than the official route?

There is no honest raw-credit comparison: Wan.video credits and GlobalGPT credits belong to different billing systems. What we can say is narrower and more useful:

1. The official Wan consumer route sells memberships and Wan credits.
2. Our completed GlobalGPT tests returned $0.70 and $1.40 for two concrete WAN 3.0 jobs, with the matching creation-surface prices shown as 700 and 1,400 credits.
3. GlobalGPT lets a creator use other video models from the same account, so a team does not need to fund a separate provider balance just to compare a WAN shot with another workflow.

That flexibility matters when the brief changes. You may start with WAN 3.0 for a reference-driven shot, then compare it with a Seedance 2.0 pricing workflow, sebuah Veo 3.1 access route, atau sebuah Gemini Omni Flash vs. Veo 3.1 test. The value claim is the shared platform workflow—not an assertion that one provider’s credits are equivalent to another’s.

Coba GlobalGPT if you want to run WAN 3.0 alongside other video models from one account.

What about official WAN 3.0 API pricing?

Do not treat the official Wan membership table as an API rate card. We confirmed the official consumer plans and credit packs above, but did not find a current official Alibaba Model Studio row that publishes a WAN 3.0 API unit price for the exact tested configuration. Until Alibaba publishes that same-field rate, this article does not convert Wan.video credits into an API price or present the GlobalGPT test price as an official Alibaba price.

If you need an API budget today, use the price returned by your own task submission, lock the exact settings in a test job, and record the date. This is especially important for longer clips: the current WAN 3.0 product overview discusses native 30-second generation, but a 30-second capability claim does not by itself establish a 30-second API bill.

A practical cost model for real projects

The cheapest clip is not always the cheapest project. A short paid test becomes expensive when a team has not decided its shot list, aspect ratio, and approval process before generating. Start by separating three budgets:

1. Exploration budget. This is the money for trying motion ideas, camera language, and subject consistency. In our test setup, one 5-second 480P run provided a clean baseline at $0.70. Use a small fixed number of attempts before a creative review instead of generating endlessly.
2. Selection budget. Keep the prompt and settings stable while comparing variants. That lets you tell whether a better result came from the model, the prompt, or simply a different duration.
3. Delivery budget. Only after a shot is chosen should you pay for a longer duration, a higher resolution, or a reference-driven version. This is where a 10-second run is useful: our matched test returned $1.40, exactly twice the five-second price.

If your project needs a sequence rather than one hero shot, plan the clips as a set. A 30-second social video may be cheaper to control as three independently approved ten-second shots than as one long first attempt—especially when a client changes the opening, CTA, or product angle late in the process. The right choice depends on continuity requirements, so run your own short test before committing to a long render.

Hidden costs to plan for

The first render is rarely the entire budget. Reserve room for:

• Prompt iterations and failed creative directions.
• Higher resolutions than the 480P configuration tested here.
• Audio-enabled, image-reference, video-reference, or editing workflows.
• Extra variants for a client approval round.
• Exports from a second model when you need a different motion style.

For comparison shopping, our Panduan Biaya Sora 2 dan Panduan Alternatif Sora 2 help frame the wider video-generation budget.

Is WAN 3.0 worth the price?

WAN 3.0 is easiest to justify when its longer-duration, audio-aware, or reference-driven workflow matches the shot you need. For low-risk exploration, use the official consumer credits or run a tightly controlled platform test first. For recurring production, compare the full cost of variants, revisions, and adjacent models—not only the cheapest first output.

Our two tests provide a clear starting estimate for one narrow setup: 480P, 16:9, text-to-video, no audio. At that setting, five seconds returned a $0.70 platform price and ten seconds returned $1.40. Treat that as a reproducible cost baseline, then rerun the test with your intended resolution and inputs before committing to a campaign.

PERTANYAAN YANG SERING DIAJUKAN

How much does WAN 3.0 cost?

Official Wan consumer plans start at $0/month. The official annual-billing plans display $5/month for Pro and $20/month for Premium, while credit packs run from $1.50 for 30 credits to $100 for 3,900 credits. Generation cost depends on the route and settings.

How much did your WAN 3.0 test cost on GlobalGPT?

Our completed 480P, 16:9, audio-disabled text-to-video test returned a GlobalGPT API platform price of $0.70 for five requested seconds and $1.40 for ten requested seconds. The outputs completed at 5.04 and 10.03 seconds.

How many GlobalGPT credits did the tests use?

The matching GlobalGPT creation surface showed 700 credits for the five-second 480P configuration and 1,400 credits for the ten-second configuration. The direct API response reports dollar price, so this article keeps the UI credit display and API price response explicitly separate.

Is $0.14 per second the official WAN 3.0 API price?

No. It is the observed GlobalGPT platform result for our two 480P text-to-video tests. It is not an official Alibaba API price, and it should not be applied to other resolutions, audio settings, or reference workflows.

Are Wan.video credits the same as GlobalGPT credits?

No. They are separate platform currencies with separate plans and terms. Do not compare their numbers directly without first comparing the underlying output settings and the platform’s own price basis.

Does the official Wan credit balance expire?

The official credits page says redeemed credits have a two-year validity period. It also says they cannot be refunded, transferred, withdrawn, or exchanged for memberships.

Can I use WAN 3.0 on GlobalGPT?

Yes. Our two completed direct API tasks used the wan3.0-video model and completed through the WanAli upstream provider. Availability and pricing can change, so check the current creation settings before submitting a production batch.

What should I test before buying a large credit pack?

Run one controlled clip with your target duration, resolution, aspect ratio, audio setting, and reference assets. Keep the prompt fixed for a second run so you can isolate which setting changes the cost or output quality.