OpenArt Arena Launches Expert Blind-Test Leaderboard; Seedance 2.5 Tops Video and Image
On September 16, AI creation platform OpenArt launched Arena, a global leaderboard for AI image and video models focused on "creative intelligence." Instead of a generic "strongest model" ranking, creative professionals vote in blind tests — no model names or logos shown, just two outputs judged by real creative standards — with separate boards for film, ads, animation, motion design, editing, and lip sync. The video overall top three are ByteDance Seedance 2.5 (11125), Alibaba Wan 3.0 (11047), and ByteDance Seedance 2.0; per egeberkina, Seedance 2.5 also leads the image board.
Confirmed
- Arena is built on expert blind testing across image and video generation models, organized by creative task type, aiming to answer "which model fits the content I want to make" rather than "which is strongest" (relay by minchoi, chrisfirst, azedai and others).
- Video overall top three: Seedance 2.5 (11125), Wan 3.0 (11047), Seedance 2.0 (score truncated in posts); Seedance 2.5 also ranks first on the image board.
- Per-category rankings diverge sharply from the overall board; azedai confirmed this after testing many models, and egeberkina found motion graphics, cinematic shots, lip sync, and ads may each have a different best model.
Unconfirmed
- MayorKingAI anticipated controversy and asked followers to verify the rankings; AidenTechAi also polled users on whether they agree with the top 3. The leaderboard's credibility still awaits community validation, and MayorKingAI's specific details appear only in an attached image.
Why it matters
- azedai argues "which AI model is strongest" is the wrong question — ask "strongest at what task" — and recommends Arena as a starting point for model selection by checking the board for your most common use case; egeberkina similarly advises comparing models by actual use rather than a single overall ranking.
- Being expert blind-test driven rather than automated-benchmark driven, Arena avoids brand halo and reputation bias, better reflects real creative production needs, and may shape how the creative industry picks models.
2026-09-16 ~ 2026-09-17 · 15 related posts
Primary sources
- OpenArt Arena launches a pro blind-test leaderboard for creative AI models — azed_ai · 2026-09-16
- [source] Creative AI leaderboard: Seedance 2.5 tops video, Wan 3.0 close behind — minchoi · 2026-09-16
- OpenArt Arena's current top AI models ranked, sparking debate — azed_ai · 2026-09-16
- OpenArt Arena ranks image and video models per task type, and the winners change by category — azed_ai · 2026-09-16
- OpenArt Arena launches expert-led leaderboard ranking image and video models by real creative work — chrisfirst · 2026-09-16
- Video model ranking sparks debate: Seedance 2.5 takes #1, Wan 3.0 #2, Seedance 2.0 #3 — Aiden_Tech_Ai · 2026-09-16
- OpenArt Arena Shows Video Model Rankings Shift Sharply by Use Case, Not One Winner — egeberkina · 2026-09-16
- OpenArt Arena ranks AI models blind and by use case—and rankings shift a lot — egeberkina · 2026-09-16
- OpenArt Arena launches blind-vote leaderboard ranking creative AI models by real job type — minchoi · 2026-09-17
6 near-duplicate retellings: minchoi · azed_ai · azed_ai · nevrekaraishwa2 · SimplyAnnisa · egeberkina