HyperBrowseComp: A Multilingual and Multimodal Stress Test for Web-Browsing Agents
We introduce HyperBrowseComp, a multilingual and multimodal browsing benchmark comprising 423 manually authored and human-validated questions across 13 languages, written by native or highly proficient speakers. Questions are designed to be extremely challenging. Each question targets a concise, publicly verifiable answer whose discovery requires locating obscure evidence, foll
Reading in Plain English mode — toggle below to switch
We introduce HyperBrowseComp, a multilingual and multimodal browsing benchmark comprising 423 manually authored and human-validated questions across 13 languages, written by native or highly proficient speakers. Questions are designed to be extremely challenging. Each question targets a concise, publicly verifiable answer whose discovery requires locating obscure evidence, foll
This benchmark highlights the current limitations of web-browsing agents, pushing the boundaries for multimodal and multilingual reasoning. Engineers working on agentic AI systems will need to address these challenges to build more robust and capable AI assistants.