whatmodel.sh / video / Kling O3 Pro
Kling O3 Pro
↗ share
+ contribute
Kling O3 Pro is a closed-weights video model from Kuaishou (Kling AI).
$ how it compares
The video models ranked either side of Kling O3 Pro, out of 139 with a score, then more from Kuaishou (Kling AI).
| # | model | score | where | ||
|---|---|---|---|---|---|
| 29 |  Wan2.1-VACE-14B | 65±7 | comparison only | ||
| 30 |  Veo 3 | 64±2 | comparison only | ||
| 31 | Kling O3 Pro | 64±4 | api $0.11/sec | ||
| 32 |  Seedance 1.0 Pro Fast | 62±11 | api $0.049/sec | ||
| 33 | Wan 2.5 Preview | 61±2 | api $0.10/sec | ||
| more from Kuaishou (Kling AI) | |||||
| 22 | Kling 3.0 | 68±4 | api $0.084/sec | ||
| 34 | Kling 2.6 Pro | 59±2 | api $0.070/sec | ||
| 44 | Kling O1 Pro | 56±2 | api $0.11/sec | ||
$ what it's good at
video editing▄··█··▄··█▄█ #8 of 9 * · 61% of its score · 1 test
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena video edit | 1255.3 ✎ | █··█··▄··█▄█ beats 2/9 | 61% |
other tests1 not in a task · 32% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| WROP object-permanence exam (human Elo) | 1471.3 ✎ | ▄····██···▄█ beats 1/7 | 32% |
$ where to run it
Through an API
api · fal.ai · $0.11/sec · fal.ai ↗ · price sourceUSD per second of generated video on fal.ai, audio off. Product: fal Kling O3 Pro text-to-video endpoint ('Kling O3 Text to Video [Pro]'). API model: fal-ai/kling-video/o3/pro/text-to-video. Settings: generate_audio=false (the endpoint default is true, $0.14/s). Price basis: per-second rate printed by fal. The O3 Pro video-edit endpoint (fal-ai/kling-video/o3/pro/video-to-video/edit, input video) is $0.168/s. fal does not state the output resolution on this page. Kling's own developer page lists 'Kling 3.0 Omni' without an O3 or Pro label, so the explicitly named fal endpoint is used. Currency: USD. Not the Kling direct tariff. fal.ai model page, observed 2026-09-26.
$ how the score was worked out · LMArena leaderboard dataset 61%, Training Object Permanence in… 32%
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 | 1 | 61% |
 Training Object Permanence in World Models (WROP exam; arXiv:2609.28654v1) | 1 | 32% |
| prior · pull toward the middle | 7% | |
| weighted average | = 63.7 |
$ 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