Benchmark
BrowseComp-zh
A high-difficulty benchmark purpose-built to comprehensively evaluate LLM agents on the Chinese web, consisting of 289 multi-hop questions spanning 11…
- Modality
- text
- Categories
- reasoning, search
- Openness
- unknown
- Source
- llm_stats
- Reported scores
- 13
Reported scores
llm_stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| DeepSeek-R1-0528 | DeepSeek | 0.357 | 2025-05-28 | evidence |
| DeepSeek-V3.1 | DeepSeek | 0.492 | 2025-01-10 | evidence |
| DeepSeek-V3.2 | DeepSeek | 0.65 | 2025-12-01 | evidence |
| DeepSeek-V3.2 (Thinking) | DeepSeek | 0.65 | 2025-12-01 | evidence |
| DeepSeek-V3.2-Exp | DeepSeek | 0.479 | 2025-09-29 | evidence |
| GLM-4.7 | Z.ai | 0.666 | 2025-12-22 | evidence |
| Kimi K2-Thinking-0905 | Moonshot AI | 0.623 | 2025-09-05 | evidence |
| LongCat-Flash-Thinking-2601 | Meituan | 0.69 | 2026-01-14 | evidence |
| MiniMax M2 | MiniMax | 0.485 | 2025-10-27 | evidence |
| Qwen3.5-122B-A10B | Qwen | 0.699 | 2026-02-24 | evidence |
| Qwen3.5-27B | Qwen | 0.621 | 2026-02-24 | evidence |
| Qwen3.5-35B-A3B | Qwen | 0.695 | 2026-02-24 | evidence |
| Qwen3.5-397B-A17B | Qwen | 0.703 | 2026-02-16 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.