Best AI video generation models, ranked by quality
Models that generate a video clip from a text prompt, compared on Arena quality, the price of one default clip and quality per dollar.
Best AI video generation models, ranked by quality
Sorted by Arena score, highest first. Models without an Arena rating follow, by price.
Median Arena score here: 1,191
| # | Model | Arena score | Price | Value | Actions |
|---|---|---|---|---|---|
| 01 | Google Veo 3.1 Google DeepMindImage-to-Video Arena: 1,390 | Arena1,362±10 · 25,076 votes | $0.48/s$1.92 for 4 s | 89 pts/$+171 over median | Try |
| 02 | Google Veo 3.1 Fast Google DeepMindImage-to-Video Arena: 1,372 | Arena1,358±10 · 25,870 votes | $0.18/s$0.72 for 4 s | 232 pts/$+167 over median | Try |
| 03 | Arena1,342±6 · 165,462 votes | $0.060/s$0.30 for 5 s | 504 pts/$+151 over median | Try | |
| 04 | Arena1,240±11 · 32,434 votes | $0.084/s$0.42 for 5 s | 116 pts/$+49 over median | Try | |
| 05 | Runway Gen 4.5 Runway | Arena1,225±9 · 48,127 votes | $0.144/s$0.72 for 5 s | 47 pts/$+34 over median | Try |
| 06 | Arena1,206±6 · 89,138 votes | $0.336/video | 45 pts/$+15 over median | Try | |
| 07 | Arena1,191±11 · 12,164 votes | $0.18/s$0.90 for 5 s | Not clearly above the median | Try | |
| 08 | Arena1,181±12 · 9,379 votes | $0.324/video | Not clearly above the median | Try | |
| 09 | Arena1,164±16 · 6,531 votes | $0.60/s$4.80 for 8 s | Not clearly above the median | Try | |
| 10 | Arena1,163±10 · 14,104 votes | $0.336/s$1.68 for 5 s | Not clearly above the median | Try | |
| 11 | Arena1,113±10 · 16,253 votes | $0.0432/s$0.216 for 5 s | Not clearly above the median | Try | |
| 12 | Arena1,065±17 · 5,250 votes | $0.216/s$1.08 for 5 s | Not clearly above the median | Try | |
| 13 | Mochi 1 Genmo | Arena1,007±17 · 5,893 votes | ≈ $0.5041/run | Not clearly above the median | Try |
| Without an Arena rating (14), by price | |||||
| – | LTX-Video (Lightricks) Lightricks | No Arena rating | ≈ $0.0229/run | Try | |
| – | Wan 2.2 5B Fast Alibaba (Wan) | No Arena rating | $0.030/video | Try | |
| – | Wan 2.2 Text-to-Video Alibaba (Wan) | No Arena rating | $0.060/video | Try | |
| – | No Arena rating | $0.060/s$0.30 for 5 s | Try | ||
| – | Luma Ray Flash 2 Luma AI | No Arena rating | $0.072/s$0.36 for 5 s | Try | |
| – | Seedance 1 Pro Fast ByteDance | No Arena rating | $0.072/s$0.36 for 5 s | Try | |
| – | Kling V2.5 Turbo Pro Kuaishou (Kling) | No Arena rating | $0.084/s$0.42 for 5 s | Try | |
| – | No Arena rating | $0.60/video | Try | ||
| – | Wan 3 Alibaba (Qwen) | No Arena rating | $0.12/s$0.60 for 5 s | Try | |
| – | No Arena rating | $0.18/s$0.72 for 4 s | Try | ||
| – | Kling v3 Kuaishou (Kling)Image-to-Video Arena: 1,354 | No Arena rating | $0.2688/s$1.344 for 5 s | Try | |
| – | Kling v3 Omni Kuaishou (Kling) | No Arena rating | $0.2688/s$1.344 for 5 s | Try | |
| – | HunyuanVideo Tencent | No Arena rating | ≈ $3.06/run | Try | |
| – | No Arena rating | $0.48/s$3.84 for 8 s | Try | ||
Value = Arena points above the median of the rated models in this list per US dollar of one default run; only models whose 95 % interval lies fully above the median get a value rank.
How to read this ranking
- Arena score
- A rating from blind pairwise votes on the public Arena: people compare two anonymous models on the same prompt and pick the better result. Higher is better; ± is the 95 % interval, “votes” the number of battles.
- Price
- What one run with the model’s default settings costs on Railwail, from the same pricing rules that bill your runs (1 credit = $0.01). ≈ marks GPU-time prices: a typical run, billed by the real run time.
- Value
- Value = Arena points above the median of the rated models in this list per US dollar of one default run; only models whose 95 % interval lies fully above the median get a value rank.
- Who is listed
- Lab models you can run on Railwail that create a result from a text prompt alone, plus every model rated on this arena. Tools such as upscalers and community uploads are on the full leaderboard.
- No Arena rating
- The arena has no entry for exactly this model in the setting we run (for example, only a high-reasoning or a 1080p run is rated). We never estimate a score.
- Rank
- The position within this list, not the rank on the arena.
Source
Quality scores: Text-to-Video Arena leaderboard, published Sep 22, 2026, retrieved Sep 24, 2026 from the LMArena leaderboard dataset. Used under CC BY 4.0; we show the overall score of the models we can match to a Railwail model and rank them within this list.
Questions about this ranking
Which AI video generation model has the highest Arena score on Railwail?
Google Veo 3.1 by Google DeepMind with an Arena score of 1,362 (Text-to-Video Arena, published Sep 22, 2026). Next: Google Veo 3.1 Fast (1,358) and Grok Imagine Video (1,342).
Which AI video generation model is the cheapest on Railwail?
LTX-Video (Lightricks) by Lightricks at ≈ $0.0229/run. Prices come from the same pricing rules that bill your runs.
Which AI video generation model gives the most quality per dollar?
Grok Imagine Video: 151 Arena points above the median of this list (1,191) at $0.060/s, that is 504 points per dollar.
How is value (quality per dollar) calculated?
Value = Arena points above the median of the rated models in this list per US dollar of one default run; only models whose 95 % interval lies fully above the median get a value rank.
Why do some models have no Arena rating?
The public Arena rates many models only in one setting, for example with high reasoning effort, with web search or at 1080p. We show a score only when the rated entry is exactly the model and default setting that runs on Railwail; otherwise the model is listed without a score.