What it measures
- Web browsing
- Deep search
- Evidence gathering
BrowseComp
A browsing benchmark for hard-to-find information and deep web research tasks.
How to read the score
% accuracy; higher is better.
Sourced Ranking
| Rank | Model | Provider | Score | Source | Source Type | Verified |
|---|---|---|---|---|---|---|
| 1 | Gemini 3.1 Pro | 85.9 | Gemini 3.1 Pro model page Accessed 2026-06-11 | official | Yes | |
| 2 | OpenAI GPT-5.4 | OpenAI | 82.7 | Introducing GPT-5.4 Accessed 2026-06-11 | official | Yes |
| 3 | OpenAI GPT-5.2 Thinking | OpenAI | 65.8 | Introducing GPT-5.2 Accessed 2026-06-11 | official | Yes |
| 4 | Gemini 3 Pro | 59.2 | Gemini 3.1 Pro model page Accessed 2026-06-11 | official | Yes |
Limitations
- Cite model-release score sources separately.
- The arXiv paper is static; latest scores usually appear in model release pages.
- OpenAI page and code are official entry points, but not a single dynamic leaderboard.
leaderboard
OpenAI BrowseComp page and simple-evals code
The arXiv paper is static; latest scores usually appear in model release pages. OpenAI page and code are official entry points, but not a single dynamic leaderboard.
paper
BrowseComp paper or specification
Use for benchmark definition, not necessarily for latest model scores.
third-party
BrowseComp dataset or code
Use for dataset, code, or implementation details; score freshness depends on the benchmark.