Short answer: Wan 3.0 does more than Wan 2.7: clips up to 30 seconds (Wan 2.7 stops at 15), and one model that takes reference images, reference audio, and even a document or a web page. In our same-prompt test, Wan 3.0 Prime followed the requested camera move more closely. On price, standard Wan 3.0 is $0.08 per second at 720P, cheaper than Wan 2.7's $0.10. Wan 3.0 Prime costs more ($0.18 at 720P) but generates faster and lists a 480P draft tier at $0.0875 per second.
The differences at a glance
| Wan 3.0 Prime | Wan 3.0 | Wan 2.7 | |
|---|---|---|---|
| Model id | wan3.0-video-prime | wan3.0-video | wan2.7-video/text-to-video, /image-to-video |
| Clip length | 2–30 s | 2–30 s | 2–15 s |
| 480P / second | $0.0875 | not listed | not available |
| 720P / second | $0.18 | $0.08 | $0.10 |
| 1080P / second | $0.36 | $0.16 | $0.167 |
| First / last frame | Yes | Yes | Yes (image-to-video) |
| Reference images | Up to 10 | Up to 10 | No |
| Reference video | No | Up to 5 clips, 15 s total | Continue from a first clip (first_clip) |
| Reference audio | Up to 5 | Up to 5 | Driving audio |
| Document or web page reference | One of each | One of each | No |
| Sound | audio, on by default | audio, on by default | Both of our clips came back with audio |
| Default resolution | 1080P | 1080P | 1080P |
Live HiAPI prices, checked October 15, 2026. Resolution values must be uppercase.
Same prompt, side by side
Both tests use the same prompt, 5 seconds, 16:9. Wan 2.7's lowest tier is 720P, while the Prime clips are at the 480P draft tier, so don't compare sharpness. Compare the camera move, framing, and how closely each followed the prompt.
Test 1: pistachio gelato with a 180-degree orbit
A single scoop of pistachio gelato in a frosted coupe glass on a pale marble
counter; slow 180-degree orbit as a bead of condensation slides down the glass,
soft window light from the left, shallow depth of field, warm late-afternoon
tones, faint steam curling from an espresso cup just behind in soft focus.
Wan 3.0 Prime:
Run this prompt on Wan 3.0 Prime →
wan3.0-video-prime, 480P, 5 seconds, no sound, $0.4375.
Wan 2.7:
wan2.7-video/text-to-video, 720P, 5 seconds, $0.50.
What's visibly different: Prime orbits, keeps the whole glass in frame, and brings the espresso cup into view as it turns. Wan 2.7 opens wide and quickly pushes into a big close-up of the scoop, which reads as a push-in rather than an orbit. Both show the condensation and the coffee steam.
Test 2: a rainy neon alley, with sound
Rain streaking down a narrow neon-lit alley at night; a slow dolly-forward past
steaming street-food stalls, magenta and cyan reflections rippling in the wet
pavement, a tabby cat darting between crates mid-shot. Ambient sound: steady
rain, distant traffic hiss, the sizzle of a grill and a low synth drone.
Wan 3.0 Prime:
Run this prompt on Wan 3.0 Prime →
wan3.0-video-prime, 480P, 5 seconds, with sound, $0.4375.
Wan 2.7:
wan2.7-video/text-to-video, 720P, 5 seconds, $0.50.
What's visibly different: both have magenta and cyan reflections on wet pavement, steam, and a forward camera move. Prime lines the alley with lit food stalls and shows the cat small, mid-shot. Wan 2.7 has fewer stalls, and its cat crosses the foreground in the second half, more prominently. Both came back with ambient sound.
Which to choose
- Clips longer than 15 seconds: Wan 3.0, up to 30 seconds. Wan 2.7 caps at 15.
- Lock a character or product with reference images, or hand the model a brand doc or product page: Wan 3.0. Wan 2.7 can't.
- Follow a reference video's motion or camera: standard Wan 3.0. Prime doesn't take reference videos.
- Lowest cost when speed doesn't matter: standard Wan 3.0, $0.08 per second at 720P, the cheapest 720P of the three.
- Fast results, or cheap 480P drafts: Wan 3.0 Prime, $0.0875 per second at 480P.
- Already integrated Wan 2.7 for text or image-to-video: it still works. If you move to Wan 3.0, the model id, aspect parameter, and image parameter all change (see below).
Moving from Wan 2.7 to Wan 3.0
| Wan 2.7 | Wan 3.0 | |
|---|---|---|
| Model id | Separate text and image endpoints | One model id |
| Aspect parameter | ratio | aspect_ratio (adds adaptive) |
| First frame | media: [{"type": "first_frame", "url": ...}] | first_frame_url |
| Sound | Text-to-video can take an audio_url to drive it | audio switch, on by default |
| Length | 2–15 s | 2–30 s |
{
"model": "wan3.0-video",
"input": {
"prompt": "A single scoop of pistachio gelato in a frosted coupe glass; slow 180-degree orbit",
"resolution": "720P",
"aspect_ratio": "16:9",
"duration": 5
}
}
FAQ
Is Wan 3.0 always more expensive than Wan 2.7?
No. Standard Wan 3.0 is $0.08 per second at 720P, less than Wan 2.7's $0.10. Prime is faster and costs $0.18 at 720P.
Do Wan 3.0 Prime and standard Wan 3.0 produce the same output?
The docs describe the same inputs and output spec. Prime generates faster, lists a 480P tier, and doesn't take reference videos.
Can I just change the model name in my Wan 2.7 request?
No. Rename ratio to aspect_ratio, and replace image-to-video's media array with first_frame_url / last_frame_url.
Code is in the Wan 3.0 Prime API guide and the Wan 2.7 API guide, and pricing in Wan 3.0 Prime API pricing.
