[AINews] Opus 5.5 is good at explainer videos

Claude Opus 5.5 landed this week, and the loudest reaction wasn’t about benchmarks or coding. It was about videos. More specifically, explainer videos made from code, the sort of thing that can make the old “AI video” playbook look a little less fixed than people assumed.

The source roundup says Opus 5.5 and GPT-6 Astra are generating videos and animations entirely from code, including a p5.brush 4K “time” film, Blender claymation skills, and a 400+ hour Astra 3D scene. That is what pushed the model into the center of the conversation. The vibe was positive enough that the model was said to have taken over the timeline for explainer videos.

This all sits inside a bigger wave of model news. Opus 5.5 now leads SimpleBench at 88.4%, and on vision it is described as Anthropic’s best model to date, ahead of Fable 5 and GPT-6 Sol but behind GPT-6 Astra, while costing about 60% less than Fable 5.1. It also moves from 24% to 62% on Terminal-Bench-Science as reasoning effort rises from low to xhigh, before slipping to 59% at max.

But the social energy here is not really about those numbers. It’s about a model that people are treating like a media engine, not just a chatbot. The source even notes the reaction was “who knew you didn’t need diffusion,” which is a pretty good summary of why this landed so hard.