Benchmark
BrowseComp
reasoning
search
text
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.
Top 10
榜单概览
数据说明
分数为 0-100 百分制综合成绩（Elo 榜为相对评分，仅供同榜内比较）。评测分数属事实数据，整理自公开评测体系与厂商公开资料，部分分数为厂商自报口径，不同榜单的评测方法与数据集可能存在差异，请结合多个榜单交叉参考。
Ranking
模型排名
| 名次 | 模型 | 机构 | 分数 | 来源 |
|---|---|---|---|---|
| 1 | GPT-5.6 Sol | OpenAI | 90.4 | 来源 |
| 2 | GPT-5.5 Pro | OpenAI | 90.1 | 来源 |
| 3 | GPT-5.4 Pro | OpenAI | 89.3 | 来源 |
| 4 | GPT-5.6 Terra | OpenAI | 87.5 | 来源 |
| 5 | Claude Sonnet 5 | Anthropic | 84.7 | 来源 |
| 6 | GPT-5.5 | OpenAI | 84.4 | 来源 |
| 7 | MiniMax-M3 | MiniMax | 83.5 | 来源 |
| 8 | GPT-5.6 Luna | OpenAI | 83.3 | 来源 |
| 9 | GPT-5.4 | OpenAI | 82.7 | 来源 |
| 10 | LongCat-2.0 | Meituan | 79.9 | 来源 |
| 11 | Step 3.7 Flash | StepFun | 75.8 | 来源 |
| 12 | GPT-5 | OpenAI | 54.9 | 来源 |
| 13 | GLM-4.6 | Zhipu AI | 45.1 | 来源 |
| 14 | Grok 4 Fast | xAI | 44.9 | 来源 |
| 15 | Nemotron 3 Ultra 550B A55B | NVIDIA | 44.4 | 来源 |
| 16 | DeepSeek-V3.2-Exp | DeepSeek | 40.1 | 来源 |
| 17 | GLM-4.5 | Zhipu AI | 26.4 | 来源 |
| 18 | GLM-4.5-Air | Zhipu AI | 21.3 | 来源 |