# BrowseComp

> OpenAI's benchmark of 1,266 hard-to-find, short-answer questions that test how persistently and creatively browsing agents search the web.

- id: `browsecomp`
- category: Information (`information`)
- url: https://arxiv.org/abs/2504.12516
- repo: https://github.com/openai/simple-evals
- status: active
- pricing: free
- license: MIT
- tags: benchmark, web-browsing, search, agents, evals
- related: GAIA (https://indexagentica.com/entries/gaia-benchmark/); WebArena (https://indexagentica.com/entries/webarena/); Mind2Web (https://indexagentica.com/entries/mind2web/); Humanity's Last Exam (https://indexagentica.com/entries/humanitys-last-exam/)
- sources: https://arxiv.org/abs/2504.12516 , https://github.com/openai/simple-evals
- added: 2026-10-19
- updated: 2026-10-19
- last verified: 2026-10-19
- featured in: Agent benchmarks compared (https://indexagentica.com/compare/agent-benchmarks/)
- maintainer: OpenAI
- submitted by: agentica-curator
- page: https://indexagentica.com/entries/browsecomp/
- json: https://indexagentica.com/api/entries/browsecomp.json

Answers are short and easy to verify against reference answers, so grading is simple; the paper frames it as analogous to programming competitions for coding agents: incomplete, but a useful measure of persistence and creativity in finding entangled information (arXiv 2504.12516). The reference implementation lives in `openai/simple-evals`, which stopped being updated for new models in July 2025 but still hosts BrowseComp, HealthBench and SimpleQA.