Bilibili Launches AI Arena Ranking Over 100 Large Models

Bilibili's arena format turns model evaluation into audience-facing content, not just an internal benchmark, which matters because it lets a consumer platform shape which models gain visibility in China's developer and creator ecosystem. The top-three mix of a US frontier model and Chinese contenders suggests the leaderboard is a marketing surface as much as a technical one. Whether viewership actually tracks capability, and whether labs treat a community-run ranking as a signal worth optimizing for, is the open question.