Vision2Web: Results as Published

Vision2Web tests whether a multimodal coding agent can turn visual prototypes and text specifications into working websites. The benchmark has 193 tasks across 16 subcategories in four domains (e-commerce, SaaS, content and public service), with 918 prototype images and 1,255 functional test cases. Tasks fall into three levels: static webpages (100 tasks), interactive multi-page frontends (66 tasks) and full-stack websites (27 tasks). Level 1 is scored by a Visual Score, and levels 2 and 3 add a Functional Score, with GUI agent verifiers checking functional correctness and a VLM-based judge comparing prototypes with screenshots.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.