whatmodel.sh / video / UniVideo
UniVideo
↗ share
+ contribute
UniVideo is an open-weights video model from Kuaishou (Kling AI), released January 2026. Its published size is 13B parameters. Licence: apache-2.0.
Catalogue identity notes
Kling Team's UniVideo unified video understanding/generation/editing model (arXiv 2510.08377; code and model released 7 Jan 2026 per the card; KlingTeam/UniVideo pinned c84ceb4). WorldReasonBench lists 'UniVideo' among open-source generators (TI2V). Single public checkpoint. Comparison-only.
Model information sources
Release date ↗ · Weights licence ↗ · Parameter count ↗
13B as printed in VideoArgus Table 2 #P column.
$ how it compares
The video models ranked either side of UniVideo, out of 139 with a score, then more from Kuaishou (Kling AI).
| # | model | score | where | ||
|---|---|---|---|---|---|
| 101 |  Runway Gen-3 Alpha | 40±3 | comparison only | ||
| 102 |  Open-Sora Plan v1.3 (I2V) | 39±7 | local | ||
| 103 | UniVideo | 39±3 | comparison only | ||
| 104 |  T2V-Turbo (MS) | 39±6 | comparison only | ||
| 105 |  HunyuanVideo | 39±3 | api price unknown +local | ||
| more from Kuaishou (Kling AI) | |||||
| 22 | Kling 3.0 | 68±4 | api $0.084/sec | ||
| 31 | Kling O3 Pro | 64±4 | api $0.11/sec | ||
| 34 | Kling 2.6 Pro | 59±2 | api $0.070/sec | ||
$ what it's good at
image to video▁▁▃▂██▄▃▃▄▂▂ #70 of 74 * · 97% of its score · 1 test
| test | result | vs the others in its run | counts |
|---|---|---|---|
| WorldReasonBench human Elo (image+text-to-video world-state prediction) | 665.0 ✎ | ▃▃··█··▃▃▃·▃ beats 1/8 | 97% |
$ where to run it
No verified API or download recorded yet · add one
Weights and documentation: huggingface.co/KlingTeam ↗
$ how the score was worked out · WorldReasonBench: Human-Align… 97%
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 |
|---|---|---|
 WorldReasonBench: Human-Aligned Stress Testing of Video Generators as Future World-State Predictors (arXiv v1) | 1 | 97% |
| prior · pull toward the middle | 3% | |
| weighted average | = 39.2 |
$ 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