whatmodel.sh / video / Wan 3.0
Wan 3.0
↗ share
+ contribute
Wan 3.0 is a closed-weights video model from Alibaba, released August 2026. Licence: proprietary.
$ how it compares
The video models ranked either side of Wan 3.0, out of 139 with a score, then more from Alibaba.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 2 | Wan 3.0 Prime | 80±7 | api $0.14/sec | ||
| 3 |  Gemini Omni Flash | 80±2 | api $0.10/sec | ||
| 4 | Wan 3.0 | 79±2 | api $0.10/sec | ||
| 5 |  Dreamina Seedance 2.0 | 79±1 | api $0.30/sec | ||
| 6 |  MiniMax H3 | 78±1 | api $0.13/sec +local | ||
| more from Alibaba | |||||
| 9 | HappyHorse 1.1 | 76±3 | api $0.14/sec | ||
| 14 | Wan 2.2 I2V Flash | 75±4 | api $0.036/sec | ||
| 16 | Wan 2.2 I2V Plus | 71±5 | api $0.10/sec | ||
$ what it's good at
text to video▆▄▃▄▆▅█▃▂▂▂▄ #5 of 68 * · 37% of its score · 1 test
Trails FLUX 3 Video, Dreamina Seedance 2.0 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LMArena video | 1476.3 ✎ | ▂▂▃▆█▆·▆▅▁▂▇ beats 37/41 | 37% |
image to video▁▂·▂▅█▅▄▅▆▃▄ #7 of 50 · 39% of its score · 3 tests
Trails FLUX 3 Video, Dreamina Seedance 2.5, MiniMax H3 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena image-to-video | 1479.8 ✎ | ▁▁▁▁▄▇▃▆▃▂▄█ beats 38/40 | 24% |
| Rapidata physics realism (I2V, with reference) | 1183.0 ✎ | ▂··▂▂▆█▆▃▂▆▄ beats 13/19 | 14% |
| Rapidata camera movement (I2V) | 1249.2 ✎ | ▄···▄▄▄█··▄█ beats 7/8 | 0.7% |
video editing▄··█··▄··█▄█ #1 of 9 * · 7.5% of its score · 1 test
Beats Gemini Omni Flash here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena video edit | 1414.1 ✎ | █··█··▄··█▄█ beats 9/9 | 7.5% |
other tests3 not in a task · 15% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Rapidata relative camera movement (I2V, alignment Elo) | 1383.4 ✎ | █···▄▄██·█▄█ beats 11/12 | 8.7% |
| Rapidata image-to-video realism (overall Elo) | 1250.0 ✎ | ▂······▃▂▃▃█ beats 6/12 | 4.7% |
| Datapoint Video Bench (overall Elo) | 1045.9 ✎ | ▃▃·▅·▃·▃··▃█ beats 7/9 | 2.1% |
$ where to run it
Through an API
api · Alibaba Cloud Model Studio · $0.10/sec · www.alibabacloud.com ↗ · price sourceUSD per second of output video, 720P, International deployment (Singapore) list price. Product: Model Studio Wan 3.0 video generation. API model: wan3.0-video (not wan3.0-video-prime, $0.14/s at 720P). Settings: resolution=720P, text-to-video (no input video). Price basis: published list price; a limited-time 30% discount was shown and is not applied. Wan 3.0 bills input video duration too (billable duration = input + output), so video-to-video costs more. 480P $0.05/s, 1080P $0.2/s. Chinese-mainland and other regions list different USD rates. Free quota excluded. Currency: USD. Alibaba Cloud Model Studio pricing page, observed 2026-09-25. Replaces the held Replicate route (conflicting README).
$ how the score was worked out · LMArena leaderboard dataset 69%, Rapidata/world-model-physics… 14%, Rapidata/relative-camera-move… 9% and 3 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 LMArena leaderboard dataset | 3 | 69% |
 Rapidata/world-model-physics published dataset-card rankings | 1 | 14% |
| Rapidata/relative-camera-movement-human-preference-900k published dataset-card rankings | 1 | 9% |
| Rapidata/image-to-video-realism published dataset-card rankings | 1 | 5% |
| Datapoint Video Bench leaderboard (datapointai/text-to-video-human-preferences-326k dataset card) | 1 | 2% |
| Rapidata/camera-movement published dataset-card rankings | 1 | 1% |
| prior · pull toward the middle | 1% | |
| weighted average | = 78.7 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json