BrowseComp
BrowseComp is a benchmark comprising 1,266 questions that challenge AI agents to persistently navigate the internet in search of hard-to-find, entangled information. The benchmark measures agents' ability to exercise persistence in information gathering, demonstrate creativity in web navigation, and find concise, verifiable answers. Despite the difficulty of the questions, BrowseComp is simple and easy-to-use, as predicted answers are short and easily verifiable against reference answers.
Progress Over Time
Interactive timeline showing model performance evolution on BrowseComp
BrowseComp Leaderboard
| Context | Cost | License | ||||
|---|---|---|---|---|---|---|
| 1 |  Anthropic | — | — | $25.00 / $125.00 | ||
| 2 |  Google | — | 1.0M | $2.50 / $15.00 | ||
| 3 | Anthropic | — | 1.0M | $5.00 / $25.00 | ||
| 4 |  OpenAI | — | 1.0M | $2.50 / $15.00 | ||
| 5 |  Zhipu AI | 754B | 200K | $1.40 / $4.40 | ||
| 6 | OpenAI | — | 400K | $21.00 / $168.00 | ||
| 7 |  ByteDance | — | — | — | ||
| 8 |  MiniMax | 230B | 1.0M | $0.30 / $1.20 | ||
| 9 | Zhipu AI | 744B | 200K | $1.00 / $3.20 | ||
| 10 |  Moonshot AI | 1.0T | 262K | $0.60 / $2.50 | ||
| 11 | Anthropic | — | 200K | $3.00 / $15.00 | ||
| 12 |  Alibaba Cloud / Qwen Team | 397B | 262K | $0.60 / $3.60 | ||
| 12 |  StepFun | 196B | 66K | $0.10 / $0.40 | ||
| 14 | OpenAI | — | 400K | $1.75 / $14.00 | ||
| 15 | Alibaba Cloud / Qwen Team | 122B | 262K | $0.40 / $3.20 | ||
| 16 | MiniMax | 230B | 1.0M | $0.30 / $1.20 | ||
| 17 | Alibaba Cloud / Qwen Team | 27B | 262K | $0.30 / $2.40 | ||
| 17 | Alibaba Cloud / Qwen Team | 35B | 262K | $0.25 / $2.00 | ||
| 19 | Moonshot AI | 1.0T | — | — | ||
| 20 |  Xiaomi | 309B | 256K | $0.10 / $0.30 | ||
| 21 |  Meituan | 560B | 128K | $0.30 / $1.20 | ||
| 22 | OpenAI | — | 400K | $1.25 / $10.00 | ||
| 23 | Zhipu AI | 358B | 205K | $0.60 / $2.20 | ||
| 24 | OpenAI | — | 200K | $1.10 / $4.40 | ||
| 25 |  DeepSeek | 685B | 164K | $0.26 / $0.38 | ||
| 25 | DeepSeek | 685B | — | — | ||
| 27 | OpenAI | — | 200K | $2.00 / $8.00 | ||
| 28 |  Sarvam AI | 105B | — | — | ||
| 29 | Zhipu AI | 357B | 131K | $0.55 / $2.19 | ||
| 30 |  xAI | — | 2.0M | $0.20 / $0.50 | ||
| 31 | MiniMax | 230B | 1.0M | $0.30 / $1.20 | ||
| 32 | Zhipu AI | 30B | 128K | $0.07 / $0.40 | ||
| 33 | DeepSeek | 685B | — | — | ||
| 34 | Sarvam AI | 30B | — | — | ||
| 35 | 120B | 262K | $0.10 / $0.50 | |||
| 36 | DeepSeek | 671B | 164K | $0.27 / $1.00 | ||
| 37 | Zhipu AI | 355B | 131K | $0.40 / $1.60 | ||
| 38 | Zhipu AI | 106B | — | — | ||
| 39 | DeepSeek | 671B | 131K | $0.50 / $2.15 |
FAQ
Common questions about BrowseComp
Sub-benchmarks
BrowseComp Long Context 128k
A challenging benchmark for evaluating web browsing agents' ability to persistently navigate the internet and find hard-to-locate, entangled information. Comprises 1,266 questions requiring strategic reasoning, creative search, and interpretation of retrieved content, with short and easily verifiable answers.
BrowseComp Long Context 256k
BrowseComp is a benchmark for measuring the ability of agents to browse the web, comprising 1,266 questions that require persistently navigating the internet in search of hard-to-find, entangled information. Despite the difficulty of the questions, BrowseComp is simple and easy-to-use, as predicted answers are short and easily verifiable against reference answers. The benchmark focuses on questions where answers are obscure, time-invariant, and well-supported by evidence scattered across the open web.
BrowseComp-VL
BrowseComp-VL is the vision-language variant of BrowseComp, evaluating multimodal models on web browsing comprehension tasks that require processing visual web page content alongside text.
BrowseComp-zh
A high-difficulty benchmark purpose-built to comprehensively evaluate LLM agents on the Chinese web, consisting of 289 multi-hop questions spanning 11 diverse domains including Film & TV, Technology, Medicine, and History. Questions are reverse-engineered from short, objective, and easily verifiable answers, requiring sophisticated reasoning and information reconciliation beyond basic retrieval. The benchmark addresses linguistic, infrastructural, and censorship-related complexities in Chinese web environments.