Kling 3.0 Omni is 9th for video editing at 1,000 Elo: Kling 3 on Sume

Kling 3.0 Omni at 1080p (Pro) is ninth on Artificial Analysis's video-editing board (AA-Video-Editing v1.1) with 1,000 Elo and a listed $10.91 per minute, which is 187 points behind Wan 3.0 in first place. Sume does not list a Kling Omni id. It lists kling-3, a generation model for 4 to 15 second clips at 720p or 1080p, billed at $0.14 per second with audio off and $0.21 with audio on.

Where Kling sits on the board

The board ranks editing a video from a text instruction with the audio kept. Kling Omni is the only Kling entry among the first ten rows.

What Sume lists for Kling

Sume's video list has one Kling generation id, kling-3. It takes a prompt and first and last frame images. It accepts no reference_image_urls or reference_video_urls, so it cannot use a source clip as a reference, and the edit route is documented only for Gemini Omni Flash 1.1. Its price in Sume's table is a flat per-second rate that does not change with resolution.

Do not compare the two prices directly

The $10.91 on the board is for a Kling Omni edit at 1080p Pro. The $12.60 on Sume is for a kling-3 generation with audio. They are different products and different jobs, and the numbers happen to be close. The comparison that does hold is Kling against the edit leaders on the same board: Kling sits 139 points below Gemini Omni Flash 1.1 (1,139), the model Sume uses for edits, at a similar listed price per minute.

What to pick

For a prompt-to-clip job where Kling's look matters, use kling-3 and send generate_audio explicitly so the audio price is not a surprise. For editing an existing clip, use the Omni edit route from the Wan 3.0 and Omni editing page. Keep in mind that a clip longer than 15 seconds does not fit kling-3, while Wan 3.0 takes 30.

How to cost a Kling job

Because kling-3 is billed at a flat rate per second whatever the resolution, the cost of a job is seconds times the audio rate. Ten seconds with audio off is 10 x $0.14 = $1.40, and with audio on 10 x $0.21 = $2.10. Fifteen seconds, the model's maximum, is $2.10 or $3.15.

Send the audio choice explicitly. Audio adds 50% to the price (0.21 / 0.14 = 1.5), so an unintended audio-on batch of twenty 10 second clips costs $42.00 instead of $28.00. Check the model's capabilities in GET /v1/videos/models before you plan the batch.

Kling's one appearance in the rows read is the editing board, where Kling 3.0 Omni sits ninth. This page does not claim it is weak at generation, because the text-to-video and image-to-video boards in the rows read do not list a Kling entry in their top 12. A model outside the top 12 may still suit a brief; the board only ranks the entries it lists.

To judge kling-3 for yourself, run one 5 second prompt at $0.70 without audio and compare it with a Wan 3.0 clip at $0.625 (720p).