It evaluates models on creating short product and brand videos in code from 114 design briefs, with the output rendered using HyperFrames HTML.
The launch results, dated October 26, 2026, cover 27 models. Claude Opus 5.5 (max with fallback) leads at 1536 Elo, 93 points ahead of the next model, and ranks first on all five judge axes. Claude Sonnet 5.5 scores 1443, GPT-6 Astra 1427, and GPT-6.1 Sol 1369.
Artificial Analysis reports per-task costs ranging from $0.01 to $3.89, a spread of more than 300x. GPT-6.1 Sol is listed at $0.31 per task, compared with $3.08 for Claude Opus 5.5. The highest-ranked open-weights models in the results trail the leader by at least 483 Elo points.
The source establishes results for this benchmark’s briefs, judges, model settings, and HTML rendering workflow. It does not establish how the same models perform on arbitrary video-production work, nor the total cost or quality of a finished commercial video.