EvoBrowseComp-ZH: leaderboard
Metric: Accuracy (%) on the 400 Chinese EvoBrowseComp questions (complex multi-hop questions synthesized from fresh live-web knowledge to resist parametric recall), search agent with web search and page visit tools (Tongyi DeepResearch tool definitions), answers judged correct against the reference by a GLM-5-Chat judge; temperature 0.6, top_p 0.95, 128K context, at most 40 tool calls; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.8 (Max) | 38.3 |
| 2 | Claude Opus 4.6 (Max) | 36.8 |
| 3 | Qwen 3.5 397B A17B | 34.5 |
| 4 | GLM-5 | 30.5 |
| 5 | DeepSeek V3.2 (Thinking) | 30.5 |
| 6 | DeepSeek V4 Flash (High) | 24.8 |
| 7 | Qwen 3.5 122B A10B | 21 |
| 8 | DeepSeek V4 Flash (Non-reasoning) | 20 |
| 9 | Kimi K2.6 | 19.8 |
| 10 | Qwen 3.5 27B | 18.6 |
| 11 | Qwen 3 235B A22B 2507 (Thinking) | 17.5 |
| 12 | DeepSeek V4 Pro (Max) | 15 |
| 13 | Qwen 3.5 35B A3B | 14.2 |
| 14 | DeepSeek V4 Flash (Max) | 10.8 |
Interactive version: theaggregate.ai/benchmark?slug=evobrowsecomp-zh · How It Works · Data refreshed daily, snapshot 2026-09-29.