Skip to main content

Compare AI video models, task by task

23 relevant exact model versions, 253 direct matchups, and two strictly separate blind-preference leaderboards for text-to-video and image-to-video.

video models
23
matchups
253
published ratings
40
separate tasks
2

Current leaderboards

AI video model leaderboards

Text-to-video and image-to-video remain separate because LMArena measures different input tasks. Each list shows the published results for its task. Missing scores are omitted instead of being estimated.

Text-to-video leaderboard

Blind preference ratings for videos generated from a written prompt.

1.Gemini Omni Flash1,511.6, Rank 1
2.FLUX.3 Video1,494.4
3.Dreamina Seedance 2.0 720p1,481.7
4.Dreamina Seedance 2.5 720p1,477
6.MiniMax H31,452.7
8.Sora 2 Pro1,364
9.Veo 3.11,363.9
11.Veo 3.1 Fast1,361.4
14.Grok Imagine Video1,345.4
15.Wan 2.71,344.4
16.Sora 21,340.1
18.Wan 2.61,329.8
23.PixVerse V5.61,239.7
24.Runway Gen-4.51,222.7
26.Kling 2.6 Pro1,216.9
30.Hailuo 2.31,204.3
35.HunyuanVideo 1.51,169.2
38.LTX-2 19B1,151.1
39.Wan 2.2 A14B1,131
44.Pika 2.21,007.7

20 of 23 comparison models have a published result. Ranks refer to the complete LMArena cohort. Data retrieved August 26, 2026.LMArena

Image-to-video leaderboard

Blind preference ratings for videos generated from a supplied still image.

1.MiniMax H31,489.4, Rank 1
2.Dreamina Seedance 2.5 720p1,483.7
3.Dreamina Seedance 2.0 720p1,478.2
4.Gemini Omni Flash1,462.2
5.Grok Imagine Video 1.51,460.3
6.FLUX.3 Video1,449.3
8.Wan 2.71,427.5
9.Grok Imagine Video1,415
10.Veo 3.11,397.3
12.Veo 3.1 Fast1,384.2
15.Vidu Q3 Pro1,362.5
16.Kling 3.0 Pro1,356
20.Wan 2.61,310.7
22.PixVerse V5.61,299.2
23.Kling 2.6 Pro1,293.1
26.Hailuo 2.31,260.7
37.HunyuanVideo 1.51,195.8
40.Wan 2.2 A14B1,169.2
42.LTX-2 19B1,154
45.Pika 2.2995.9

20 of 23 comparison models have a published result. Ranks refer to the complete LMArena cohort. Data retrieved August 26, 2026.LMArena

Quick start

The 20 most popular model comparisons

Here you'll find the 20 matchups that matter right now. Compare general data, prices, benchmarks, and more.

  1. Sora 2 vs. Veo 3.1
  2. Veo 3.1 vs. Runway Gen-4.5
  3. Dreamina Seedance 2.0 720p vs. Kling 3.0 Pro
  4. Gemini Omni Flash vs. MiniMax H3
  5. Veo 3.1 vs. Grok Imagine Video 1.5
  6. MiniMax H3 vs. Wan 2.7
  7. Gemini Omni Flash vs. FLUX.3 Video
  8. Gemini Omni Flash vs. Dreamina Seedance 2.5 720p
  9. Gemini Omni Flash vs. Sora 2 Pro
  10. Gemini Omni Flash vs. Veo 3.1
  11. Gemini Omni Flash vs. Veo 3.1 Fast
  12. Gemini Omni Flash vs. Grok Imagine Video 1.5
  13. FLUX.3 Video vs. Dreamina Seedance 2.5 720p
  14. Dreamina Seedance 2.5 720p vs. Sora 2 Pro
  15. Dreamina Seedance 2.5 720p vs. Veo 3.1
  16. MiniMax H3 vs. Sora 2 Pro
  17. MiniMax H3 vs. Veo 3.1
  18. Sora 2 Pro vs. Sora 2
  19. Sora 2 Pro vs. Grok Imagine Video 1.5
  20. Veo 3.1 vs. Kling 3.0 Pro

Selection

The 23 most important video models

The list focuses on current leaders, relevant lower-cost variants, and open-weight alternatives with published benchmark coverage.

For the full technical profile of every model, browse the AI model directory.

Google

Gemini Omni FlashVeo 3.1Veo 3.1 Fast

Black Forest Labs

FLUX.3 Video

ByteDance

Dreamina Seedance 2.5 720pDreamina Seedance 2.0 720p

MiniMax

MiniMax H3Hailuo 2.3

OpenAI

Sora 2 ProSora 2

xAI

Grok Imagine Video 1.5Grok Imagine Video

Alibaba

Wan 2.7Wan 2.6Wan 2.2 A14B

Runway

Runway Gen-4.5

Kuaishou

Kling 3.0 ProKling 2.6 Pro

ShengShu

Vidu Q3 Pro

PixVerse

PixVerse V5.6

Lightricks

LTX-2 19B

Tencent

HunyuanVideo 1.5

Pika

Pika 2.2

Frequently asked questions about video models

Answers about benchmark scope, pricing, and comparable video samples.

Changelog

The latest updates and improvements to our AI video model comparison
v1.2August 28, 2026

Video model leaderboards

  • Added separate text-to-video and image-to-video leaderboards directly below the hero
  • Showed published source results for each task
  • Omitted missing source results instead of estimating ratings or ranks
v1.1August 24, 2026

Controlled sample-video protocol

  • Added a fixed prompt, one-request and first-output rule
  • Normalized playback to five seconds, 16:9, 1,280 × 720 pixels, 30 fps, and muted audio without cropping
  • Stored source metadata plus original and normalized SHA-256 hashes
v1.0August 24, 2026

Initial release

  • Published 23 relevant exact model versions and 253 canonical matchups
  • Integrated separate LMArena text-to-video and image-to-video dataset revisions
  • Added pricing, specifications, provider branding, methodology, FAQ, and navigation