Claude Opus 5.5 gains traction for motion-graphics video generation and leads OpenRouter spend

Claude Opus 5.5 shipped this week and quickly gained attention for its ability to generate motion-graphics explainer videos from a single prompt, with users reporting outputs created entirely in code and without traditional design tools like After Effects. Social posts and threads highlighted multiple examples of videos produced by Opus 5.5 under prompts such as a dynamic 15-second motion-graphics showreel and a 90-second motion design with composed piano score.

OpenRouter reported that within a week of launch, Opus 5.5 became the top Anthropic model by share of spend and share of tokens, indicating rapid user adoption and workload migration from earlier Opus versions.

On Terminal-Bench-Science, Opus 5.5’s performance varies by effort level: 24% at low effort, 62% at extra-high effort, and 59% at max effort. Community commentary recommends avoiding the max setting due to imposed minimum reasoning budgets.

On vision evaluations, Opus 5.5 is ranked as Anthropic’s best vision model to date by an external evaluator, outperforming Fable 5 and GPT-6 Sol but trailing GPT-6 Astra, while costing about 60% less than Fable 5.1.

SimpleBench places Opus 5.5 at 88.4%, positioning it as a leading frontier model on that benchmark.