Artificial Analysis Articles·· 21 小时前AI 评分48
Code2Video-AA：AI 模型用代码设计视频的能力如何？
Code2Video-AA: how well do AI models design videos in code?
AI 导读
Artificial Analysis 推出 Code2Video-AA 基准，用 Elo 评分衡量 AI 模型通过代码设计视频的能力，Claude Opus 5.5 在全部五个评估维度上领先。各模型单任务成本差距超过 300 倍，但所有模型至少能成功渲染 96% 的视频。
正文
Overview
Example tasks
Three tasks from the public split of Code2Video Bench, each with every model's rendered video. Models are in order of Code2Video-AA Elo.
How Code2Video-AA works
Claude Opus 5.5 leads on all five axes
Cost per task varies more than 300x
Every model renders at least 96% of its videos
Code2Video-AA resources
来源：Artificial Analysis Articles · artificialanalysis.ai