BrowseComp-zh: leaderboard
Chinese-language BrowseComp benchmark for hard web-search questions, testing persistent browsing and evidence synthesis over Chinese internet sources.
Metric: Score (%). Source: arxiv.org. Source: llm-stats.com. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seed 1.8 | 81.3 | #139 |
| 2 | Qwen 3.5 397B A17B | 70.3 | #146 |
| 3 | Qwen 3.5 122B A10B | 69.9 | #173 |
| 4 | Qwen 3.5 35B A3B | 69.5 | #256 |
| 5 | LongCat-Flash-Thinking-2601 | 69 | #229 |
| 6 | GLM-4.7 | 66.6 | #192 |
| 7 | DeepSeek V3.2 | 65 | #205 |
| 8 | DeepSeek V3.2 (Thinking) | 65 | #205 (DeepSeek V3.2) |
| 9 | Qwen 3.5 27B | 62.1 | #187 |
| 10 | Qwen 3 Max (Thinking) | 60.9 | #208 (Qwen 3 Max) |
| 11 | DeepSeek V3.1 | 49.2 | #264 |
| 12 | MiniMax-M2 | 48.5 | #317 |
| 13 | DeepSeek V3.2 Exp | 47.9 | #234 |
| 14 | DeepSeek R1 0528 | 35.7 | #222 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=browsecomp-zh · How It Works · Data refreshed daily, snapshot 2026-10-20.