reasoning benchmark
BrowseComp-zh Leaderboard
A high-difficulty benchmark purpose-built to comprehensively evaluate LLM agents on the Chinese web, consisting of 289 multi-hop questions spanning 11 diverse domains including Film & TV, Technology, Medicine, and History. Questions are reverse-engineered from short, objective, and easily verifiable answers, requiring sophisticated reasoning and information reconciliation beyond basic retrieval. The benchmark addresses linguistic, infrastructural, and censorship-related complexities in Chinese web environments.
Updated Aug 17, 2026
Models13
Model coverage13
MetricScore
EvidenceB
BrowseComp-zh Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAC | Score70.3% | Percentile100.0% | Participants13 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score69.9% | Percentile91.7% | Participants13 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score69.5% | Percentile83.3% | Participants13 | EvidenceC | Evaluated |
| Rank04 | ModelME | Score69.0% | Percentile75.0% | Participants13 | EvidenceC | Evaluated |
| Rank05 | ModelZA | Score66.6% | Percentile66.7% | Participants13 | EvidenceC | Evaluated |
| Rank06 | ModelDE | Score65.0% | Percentile58.3% | Participants13 | EvidenceC | Evaluated |
| Rank07 | ModelDE | Score65.0% | Percentile50.0% | Participants13 | EvidenceC | Evaluated |
| Rank08 | ModelMA | Score62.3% | Percentile41.7% | Participants13 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score62.1% | Percentile33.3% | Participants13 | EvidenceC | Evaluated |
| Rank10 | ModelDE | Score49.2% | Percentile25.0% | Participants13 | EvidenceC | Evaluated |
| Rank11 | ModelMI | Score48.5% | Percentile16.7% | Participants13 | EvidenceC | Evaluated |
| Rank12 | ModelDE | Score47.9% | Percentile8.3% | Participants13 | EvidenceC | Evaluated |
| Rank13 | ModelDE | Score35.7% | Percentile0.0% | Participants13 | EvidenceC | Evaluated |
BrowseComp-zh Highlights
The leading models and scores on this benchmark.
BrowseComp-zh Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for BrowseComp-zh
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis browsecomp-zh AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01ACQwen3.5-397B-A17BAlibaba Cloud / Qwen Team
- Score
- 70.3%
- Price
- $0.60 input / $3.6 output per 1M tokens
Strengths
- Ranks #1 of 13 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp-zh, not total model capability
- 02ACQwen3.5-122B-A10BAlibaba Cloud / Qwen Team
- Score
- 69.9%
- Price
- $0.40 input / $3.2 output per 1M tokens
Strengths
- Ranks #2 of 13 compared models
- 92th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp-zh, not total model capability
- 03ACQwen3.5-35B-A3BAlibaba Cloud / Qwen Team
- Score
- 69.5%
- Price
- $0.25 input / $2.0 output per 1M tokens
Strengths
- Ranks #3 of 13 compared models
- 83th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp-zh, not total model capability
- 04MELongCat-Flash-Thinking-2601Meituan
- Score
- 69.0%
- Speed
- Up to 100 tok/s via Meituan
Strengths
- Ranks #4 of 13 compared models
- 75th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp-zh, not total model capability
- 05ZAGLM-4.7Zhipu AI
- Score
- 66.6%
- Price
- $0.60 input / $2.2 output per 1M tokens
Strengths
- Ranks #5 of 13 compared models
- 67th percentile on this benchmark
- C evidence result
Considerations
- This result measures BrowseComp-zh, not total model capability
Selection summary
Best AI Models for BrowseComp-zh
Qwen3.5-397B-A17B currently leads BrowseComp-zh with 70.3%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is BrowseComp-zh?
What BrowseComp-zh measures and how its scores work.
A high-difficulty benchmark purpose-built to comprehensively evaluate LLM agents on the Chinese web, consisting of 289 multi-hop questions spanning 11 diverse domains including Film & TV, Technology, Medicine, and History. Questions are reverse-engineered from short, objective, and easily verifiable answers, requiring sophisticated reasoning and information reconciliation beyond basic retrieval. The benchmark addresses linguistic, infrastructural, and censorship-related complexities in Chinese web environments.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- BrowseComp-zh
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- browsecomp-zh|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about BrowseComp-zh.
Which model scores highest on BrowseComp-zh?
Qwen3.5-397B-A17B is currently ranked first with 70.3%.
What does BrowseComp-zh measure?
A high-difficulty benchmark purpose-built to comprehensively evaluate LLM agents on the Chinese web, consisting of 289 multi-hop questions spanning 11 diverse domains including Film & TV, Technology, Medicine, and History. Questions are reverse-engineered from short, objective, and easily verifiable answers, requiring sophisticated reasoning and information reconciliation beyond basic retrieval. The benchmark addresses linguistic, infrastructural, and censorship-related complexities in Chinese web environments.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
13 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.