benchmarks.wiki / Public workspace

BrowseComp

Agents / Search

BrowseComp evaluates web browsing and the ability to identify answers from multiple clues. This catalog supports introductions and discussion of search methods without republishing decrypted questions or answers.

browsecomp

Official source

History

OpenAI introduced BrowseComp in 2025 to test agents on difficult web searches whose answers must be identified from multiple clues. The benchmark targets the persistence and information gathering needed to locate obscure facts, rather than only the ability to answer familiar factual questions from memory.

At a glance

What a problem looks like

Web-browsing questions that require identifying an answer from multiple clues.

Reported scores

Evaluation conditions differ between rows, so this is not a ranking. Rows marked vendor are reports by the model’s own organization.

ModelScoreMetricSetupReported bySource
o1-2024-12-17OpenAI9.9Accuracy (%)Without browsing; medium reasoning effort.paperTable 3; Section 4.1 model identifiers
gpt-4o-search-preview-2025-03-11OpenAI1.9Accuracy (%)With browsing.paperTable 3; Section 4.1 model identifiers
gpt-4o-2024-08-06OpenAI0.6Accuracy (%)Without browsing.paperTable 3; Section 4.1 model identifiers

Why it is hard

Requires finding information on the web and combining multiple clues.

Limitations and criticism

A BrowseComp result reflects the browsing system, available tools and search budget in addition to the model. Results obtained without browsing therefore measure a different setup from results obtained with web access. Its focus on identifying answers from clues does not directly measure broader browser tasks such as completing transactions or editing documents.

How much room is left

The paper results recorded here report 9.9% accuracy for o1 without browsing at medium reasoning effort, 1.9% for GPT-4o search preview with browsing, and 0.6% for GPT-4o without browsing. These particular results are far from the ceiling and show considerable remaining difficulty for the evaluated systems. They are not a current leaderboard maximum, and their different browsing setups prevent treating them as a controlled comparison of model capability alone.

Discussion

1 thread and 0 posts on this page. Anyone may reply; no account is required.

Contribute

Share a reproduction, report an evaluation issue, or suggest a correction to this benchmark overview.

Public problem summary

Catalogue metadata only; public problem statistics are not available.

Public problems
—
Formats
—
Problems with images
—
Answer availability
—
Configs
—
Splits
—
Awaiting a first discussion
—
Problems with a discussion
—
Public contributions
—