// BENCHMARK

BrowseComp-Zh benchmark

AI model leaderboard for the BrowseComp-Zh benchmark. Compare how large language models score on BrowseComp-Zh, see the full ranking, and understand what this AI benchmark measures. DeepSeek-V3.2 currently leads with 65. Chinese-language counterpart to BrowseComp — agentic web-browsing tasks requiring multi-page information synthesis.

Leaderboard

# Model Organization Score Variant Source
#1 DeepSeek-V3.2 DeepSeek 65 thinking official ↗
#2 GPT-5 OpenAI 63 cited-deepseek-v3-2 official ↗
#3 Kimi K2 Thinking Moonshot 62.3 thinking-cited-deepseek-v3-2 official ↗
#4 Kimi K2 Thinking Moonshot 62.3 thinking-w-tools official ↗
#5 Tongyi DeepResearch Alibaba-NLP 58.1 heavy-mode official ↗
#6 o3 OpenAI 58.1 react-cited-tongyi official ↗
#7 DeepSeek-V3.1 DeepSeek 49.2 react-cited-tongyi official ↗
#8 MiniMax-M2 MiniMax 48.5 cited-deepseek-v3-2 official ↗
#9 DeepSeek-V3.2-Exp DeepSeek 47.9 from-hf-readme official ↗
#10 Tongyi DeepResearch Alibaba-NLP 46.7 react-avg@3 official ↗
#11 Claude Sonnet 4.5 Anthropic 42.4 cited-deepseek-v3-2 official ↗
#12 GLM 4.5 Zhipu AI 37.5 react-cited-tongyi official ↗
#13 Claude Sonnet 4 Anthropic 29.1 react-cited-tongyi official ↗
#14 Kimi K2 Moonshot 28.8 react-cited-tongyi official ↗
#15 Kimi K2 Moonshot 22.2 k2-0905-w-tools-cited-k2-thinking official ↗

Frequently asked questions about BrowseComp-Zh

What is the BrowseComp-Zh benchmark?

Chinese-language counterpart to BrowseComp — agentic web-browsing tasks requiring multi-page information synthesis.

How is the BrowseComp-Zh benchmark scored?

BrowseComp-Zh is scored using the accuracy metric, where a higher score is better. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.

Which AI model scores highest on BrowseComp-Zh?

As of the latest reported scores on GenAIList, DeepSeek-V3.2 achieves the highest result on BrowseComp-Zh with a score of 65.

Is a higher BrowseComp-Zh score better?

Yes. On BrowseComp-Zh a higher score indicates better performance, so models near the top of the leaderboard are the strongest.