MODEL LEADERBOARDS · AGGREGATED

Where models stand.

Across the field.

The frontier and the open-source field, ranked across the benchmarks that matter — VBench, Artificial Analysis arenas, MMMU/MMMU-Pro, OpenVLM. Refreshed regularly.

29MODELS
6SOURCES
3CATEGORIES

Text-to-video models ranked by user-preference Elo on the Artificial Analysis Video Arena, with open-weight standings from the same arena and VBench overall scores for cross-reference.

Artificial Analysis · Text-to-Video ArenaElo · as of 2026-04-30VBench LeaderboardVBench % · as of 2026-04-17
RANKMODELPROVIDERMETRICSCORE
01HappyHorse-1.0Anonymous top arena entry; Alibaba-confirmed authorship, no public API yet.AlibabaArena Elo1368 Elo
02Dreamina Seedance 2.0 (720p)Native audio + phoneme-level lip-sync.ByteDanceArena Elo1271 Elo
03Kling 3.0 1080p (Pro)Native 4K, strong multi-shot consistency.KuaishouArena Elo1248 Elo
04SkyReels V4Skywork AIArena Elo1239 Elo
05Kling 3.0 Omni 1080p (Pro)KuaishouArena Elo1233 Elo
06grok-imagine-videoxAIArena Elo1231 Elo
07Vidu Q3 ProViduArena Elo1221 Elo
08Runway Gen-4.5Pro-workflow control; the production cinematic baseline.RunwayArena Elo1216 Elo
09Veo 3.1Native synchronized audio; strongest scene consistency.Google DeepMindArena Elo1208 Elo
10Wan 2.6OPENTop open-weights model on the arena; Apache-licensed.AlibabaArena Elo1189 Elo
11LTX-2 ProOPENLightricksArena Elo1129 Elo
12LTX-2.3 FastOPENTop open-weights with native audio.LightricksArena Elo1121 Elo

Aggregated by Infer. Rankings move fast — corrections to team@infer.sh. Want to run any of these in production? Get started free →