Code2Video-AA Crowns Claude Opus 5.5

Code2Video-AA shows that generating valid HTML is table stakes; the real frontier is composition, timing, and visual taste. The leaderboard is a strong signal for agentic motion-design ability, but not a universal ranking of video-generation quality.

• –Models generate HTML and GSAP compositions, which HyperFrames renders into deterministic video.
• –Opus 5.5 ranks first across all five judging axes, with its largest advantage in craft.
• –GPT-6 Astra and GPT-6.1 Sol are competitive, while Gemini 4 Argon shows progress but remains 277 Elo points behind Opus.
• –Cost efficiency is a major differentiator: GPT-6.1 Sol costs roughly one-tenth as much per task as Opus 5.5.
• –Because Gemini 4 Argon also serves as the benchmark judge, results should be treated as directional rather than definitive.