whatmodel.sh / video / SkyReels V4
SkyReels V4
↗ share
+ contribute
SkyReels V4 is a closed-weights video model from Skywork AI, released February 2026. Licence: proprietary.
Catalogue identity notes
Comparison identity. Skywork AI's own technical report 'SkyReels-V4: Multi-modal Video-Audio Generation, Inpainting and Editing model' (arXiv 2602.21818, v1 2026-02-25, v3 2026-03-18; affiliation Skywork AI, project leader at kunlun-inc.com): joint video+audio generation, inpainting and editing, up to 1080p, 32 FPS, 15 s. The report places V4 on Artificial Analysis's video arena as of 2026-03-18, so the model was publicly served by then. No first-party API, pricing or availability page could be read: skyreels.ai is a JavaScript app with no static content, and API docs found only on resellers (Runware, WaveSpeed, EachLabs, ModelsLab). api left null. Corresponds by exact name to Arena 'SkyReels V4' (org 'skyreel', Proprietary) on text_to_video, image_to_video and video_edit. Arena does not say which endpoint or version it serves. Licence: No weights found: Skywork's Hugging Face org lists SkyReels-V3 checkpoints (A2V-19B, R2V-14B, V2V-14B) but nothing for V4 (HF API, 2026-11-12). The report (CC BY 4.0) announces no weight release.
Model information sources
$ how it compares
The video models ranked either side of SkyReels V4, out of 141 with a score.
| # | model | score | where |
|---|---|---|---|
| 9 | 76±2 | api $0.080/sec | |
| 10 | 76±2 | api $0.17/sec | |
| 11 | 76±2 | comparison only | |
| 12 | 76±3 | api $0.14/sec | |
| 13 | 76±4 | api $0.080/sec |
$ what it's good at
text to video▆▅▃█▆▇▆▂▅▁▄▃ #9 of 97 * · 55% of its score · 1 test
Beats MiniMax H3 here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LMArena video | 1458.7 ✎ | ▂▂▃▆█▃·▇▂▁█▃ beats 36/44 | 55% |
image to video▁▁▂▂█▇▅▃▃▃▃▃ #8 of 76 * · 32% of its score · 1 test
Beats Wan 3.0, Grok Imagine Video 1.5 here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena image-to-video | 1475.3 ✎ | ▁▁▁▁▄▇▃▇▂▃▅█ beats 36/42 | 32% |
video editing▄··▂▂·▂··█▂▄ #7 of 12 * · 12% of its score · 1 test
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena video edit | 1363.5 ✎ | ▄··▂▂·▂··█▂▄ beats 5/11 | 12% |
$ where to run it
No verified API or download recorded yet · add one
$ how the score was worked out · LMArena leaderboard dataset 99%
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 |
|---|---|---|
| 3 | 99% | |
| prior · pull toward the middle | 1% | |
| weighted average | = 75.9 |
$ 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