Live proxy benchmark
Subtitle-first translation stress test
Selected Models
0
Dialogue Lines
0
Completed Calls
0
Run Progress
Idle还没开始。
Summary
跑完 benchmark 后，这里会给出最快、最准、最适合实时字幕的模型。
History Ranking
Completed runs saved in this browser will accumulate here.
No saved benchmark history yet.
Leaderboard
综合分 = 准确度 45% + 可用性 35% + 延迟 20%请先选择模型并运行 benchmark。
Model Details
每条台词的原文、译文、延迟、thinking 泄漏、finish reason 都会列出来。暂无结果。