reasoning benchmark
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.
Updated Aug 7, 2026
Published models58
Registry coverage58
MetricScore
EvidenceB
BrowseComp leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
Score distribution
Top published rows on the benchmark's original scale.
BrowseComp highlights
The top published results on this benchmark's own scale.
What is BrowseComp?
Definition and scoring fields from the benchmark registry.
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.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- BrowseComp
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- browsecomp|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about BrowseComp.
Which model scores highest on BrowseComp?
Kimi K3 is currently ranked first with 91.2%.
What does BrowseComp measure?
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.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
58 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is marked as eligible for the current LLMBoard capability methodology.