Best AI for video generation
Human preference over generated clips, from a prompt, from a still, and from an edit.
Aldena runs these models inside your team's projects. See what each one costs.
| rank | model | vendor | composite | benchmarks | elo | elo | elo |
|---|---|---|---|---|---|---|---|
| 1 | Gemini Omni 1.2 Flashgoogle/gemini-omni-1.2-flash | 76.1 | 3 of 3 benchmarks | 1540 | 1503 | 1396 | |
| 2 | Wan 3.0alibaba/wan-3.0 | Alibaba | 74.9 | 3 of 3 benchmarks | 1476 | 1482 | 1424 |
| 3 | Dreamina Seedance 2.5 720pbytedance/dreamina-seedance-2.5-720p | ByteDance | 74.8 | 3 of 3 benchmarks | 1479 | 1493 | 1417 |
| 4 | Gemini Omni 1.1 Flashgoogle/gemini-omni-1.1-flash | 72.8 | 2 of 3 benchmarks | 1513 | 1487 | ||
| 5 | Minimax H3minimax/minimax-h3 | MiniMax | 69.0 | 3 of 3 benchmarks | 1456 | 1493 | 1383 |
| 6 | Gemini Omni Flashgoogle/gemini-omni-flash | 68.3 | 3 of 3 benchmarks | 1509 | 1463 | 1378 | |
| 7 | Grok Imagine Video 1.5x-ai/grok-imagine-video-1.5 | SpaceXAI | 68.0 | 2 of 3 benchmarks | 1488 | 1454 | |
| 8 | FLUX 3 Videoblack-forest-labs/flux-3-video | Black Forest Labs | 68.0 | 2 of 3 benchmarks | 1491 | 1452 | |
| 9 | Dreamina Seedance 2.0 720pbytedance/dreamina-seedance-2.0-720p | ByteDance | 65.4 | 3 of 3 benchmarks | 1483 | 1479 | 1370 |
| 10 | Hidream O1 Video 1.0hidream/hidream-o1-video-1.0 | HiDream | 61.3 | 1 of 3 benchmarks | 1455 | ||
| 11 | Muse Videometa/muse-video | Meta | 60.6 | 1 of 3 benchmarks | 1457 | ||
| 12 | Veo 3.1google/veo-3.1 | 60.4 | 2 of 3 benchmarks | 1363 | 1399 | ||
| 13 | Veo 3.1 Fastgoogle/veo-3.1-fast | 59.2 | 2 of 3 benchmarks | 1361 | 1386 | ||
| 14 | HappyHorse 1.0alibaba/happyhorse-1.0 | Alibaba | 58.7 | 3 of 3 benchmarks | 1429 | 1443 | 1305 |
| 15 | Sora 2 Proopenai/sora-2-pro | OpenAI | 58.3 | 1 of 3 benchmarks | 1367 | ||
| 16 | Wan 2.7alibaba/wan-2.7 | Alibaba | 58.1 | 2 of 3 benchmarks | 1337 | 1428 | |
| 17 | Veo 3 Fastgoogle/veo-3-fast | 56.3 | 2 of 3 benchmarks | 1347 | 1327 | ||
| 18 | Sora 2openai/sora-2 | OpenAI | 55.2 | 1 of 3 benchmarks | 1342 | ||
| 19 | Veo 3google/veo-3 | 55.1 | 2 of 3 benchmarks | 1340 | 1331 | ||
| 20 | Vidu Q3 Proshengshu/vidu-q3-pro | Shengshu | 54.9 | 1 of 3 benchmarks | 1363 | ||
| 21 | Kling v3 Prokwaivgi/kling-v3-pro | Kling AI | 54.1 | 1 of 3 benchmarks | 1353 | ||
| 22 | Wan2.6 T2Ialibaba/wan2.6-t2i | Alibaba | 52.2 | 2 of 3 benchmarks | 1324 | 1309 | |
| 23 | Wan2.5 I2I Previewalibaba/wan2.5-i2i-preview | Alibaba | 51.6 | 2 of 3 benchmarks | 1244 | 1322 | |
| 24 | Seedance 1.5 Probytedance/seedance-1-5-pro | ByteDance | 51.0 | 2 of 3 benchmarks | 1256 | 1307 | |
| 25 | Grok Imagine Videox-ai/grok-imagine-video | SpaceXAI | 50.9 | 3 of 3 benchmarks | 1341 | 1411 | 1250 |
| 26 | Pixverse v5.6pixverse/pixverse-v5.6 | PixVerse | 49.2 | 2 of 3 benchmarks | 1238 | 1299 | |
| 27 | Gen-4.5runway/gen-4.5 | Runway | 49.0 | 1 of 3 benchmarks | 1223 | ||
| 28 | Kling 2.6 Prokwaivgi/kling-2.6-pro | Kling AI | 46.9 | 2 of 3 benchmarks | 1216 | 1294 | |
| 29 | Kling 2.5 Turbo 1080pkwaivgi/kling-2.5-turbo-1080p | Kling AI | 46.9 | 2 of 3 benchmarks | 1220 | 1276 | |
| 30 | Vidu Q2 Turboshengshu/vidu-q2-turbo | Shengshu | 45.4 | 1 of 3 benchmarks | 1244 | ||
| 31 | Hailuo 2.3minimax/hailuo-2.3 | MiniMax | 44.0 | 2 of 3 benchmarks | 1205 | 1263 | |
| 32 | Kling O3 Prokwaivgi/kling-o3-pro | Kling AI | 43.5 | 1 of 3 benchmarks | 1263 | ||
| 33 | Kling v2.1 Standardkwaivgi/kling-v2.1-standard | Kling AI | 43.0 | 1 of 3 benchmarks | 1230 | ||
| 34 | Seedance v1 Probytedance/seedance-v1-pro | ByteDance | 42.8 | 2 of 3 benchmarks | 1192 | 1274 | |
| 35 | Kandinsky 5.0 T2V Prosber/kandinsky-5.0-t2v-pro | Sber AI | 42.0 | 1 of 3 benchmarks | 1175 | ||
| 36 | Ray 3luma/ray-3 | Luma AI | 41.5 | 2 of 3 benchmarks | 1206 | 1227 | |
| 37 | Hailuo 02 Prominimax/hailuo-02-pro | MiniMax | 41.0 | 2 of 3 benchmarks | 1199 | 1230 | |
| 38 | Vidu Q2 Proshengshu/vidu-q2-pro | Shengshu | 40.6 | 1 of 3 benchmarks | 1224 | ||
| 39 | Hailuo 02 Fastminimax/hailuo-02-fast | MiniMax | 39.1 | 1 of 3 benchmarks | 1195 | ||
| 40 | Kling v2.1 Masterkwaivgi/kling-v2.1-master | Kling AI | 38.1 | 2 of 3 benchmarks | 1163 | 1236 | |
| 41 | Hailuo 02 Standardminimax/hailuo-02-standard | MiniMax | 38.0 | 2 of 3 benchmarks | 1181 | 1225 | |
| 42 | Kandinsky 5.0 T2V Litesber/kandinsky-5.0-t2v-lite | Sber AI | 36.6 | 1 of 3 benchmarks | 1113 | ||
| 43 | Kling O1 Prokwaivgi/kling-o1-pro | Kling AI | 36.3 | 2 of 3 benchmarks | 1201 | 1190 | |
| 44 | Soraopenai/sora | OpenAI | 35.8 | 1 of 3 benchmarks | 1069 | ||
| 45 | Hunyuan Video 1.5tencent/hunyuan-video-1.5 | Tencent | 35.7 | 2 of 3 benchmarks | 1168 | 1199 | |
| 46 | Runway Gen4 Turborunway/runway-gen4-turbo | Runway | 34.3 | 1 of 3 benchmarks | 1054 | ||
| 47 | Mochi v1genmo/mochi-v1 | Genmo | 33.5 | 1 of 3 benchmarks | 1007 | ||
| 48 | Runway Gen4 Alephrunway/runway-gen4-aleph | Runway | 33.5 | 1 of 3 benchmarks | 1181 | ||
| 49 | Veo 2google/veo-2 | 32.7 | 2 of 3 benchmarks | 1165 | 1167 | ||
| 50 | Seedance v1 Litebytedance/seedance-v1-lite | ByteDance | 31.6 | 2 of 3 benchmarks | 1113 | 1186 | |
| 51 | Wan v2.2 A14balibaba/wan-v2.2-a14b | Alibaba | 31.6 | 2 of 3 benchmarks | 1132 | 1171 | |
| 52 | Ltx 2 19Blightricks/ltx-2-19b | Lightricks | 31.0 | 2 of 3 benchmarks | 1156 | 1161 | |
| 53 | Ray2luma/ray2 | Luma AI | 27.5 | 2 of 3 benchmarks | 1066 | 1110 | |
| 54 | Pika v2.2pika/pika-v2.2 | Pika | 25.7 | 2 of 3 benchmarks | 1009 | 998 |
How this ranks
Every benchmark value becomes a percentile among the models that have it, so accuracy scores, Elo ratings and word error rates compare without hand-tuned scaling. Metrics where lower is better are inverted first. Raw values are never summed or averaged across benchmarks. A model's mean percentile is then shrunk toward the mean of the models that were broadly benchmarked, so a model tested twice cannot outrank a broadly tested one on two lucky results. Turning a data source off runs that same ranking code again in your browser over the sources you left on.
A model scored on fewer than 2 of the 3 ranked benchmarks in this category still ranks here, on the benchmarks it does have, and its row carries a partial coverage mark. On an equal score it sits under the model that earned the same number across more of the board.
Data sources
Turn a source off to drop every benchmark it feeds and re-rank the board in your browser. Your choice follows you across the leaderboard pages.
- LMArenaCC BY 4.0
Arena ratings by LMArena, from the public leaderboard dataset.