Benchmark
BrowseComp-Plus
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.
- Models measured
- 19
- Scores recorded
- 24
- Sources
- 3
- IQ weight
- 1.07
Score distribution
- 59.6%
- Mean
- 65.9%
- Median
- 22.2%
- Std dev
- 29.5%
- 10th percentile
- 90.5%
- 90th percentile
- 10.7%
- Min
This benchmark has a weight of 1.07 for chat models, measured over 19 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | GPT 5· BrowseComp-Plus data | 95.2% |
| 2 | GLM 5.1· BrowseComp-Plus data | 90.7% |
| 3 | GPT 5.5· BrowseComp-Plus data | 90.5% |
| 4 | Claude Opus 4.5launch-page footnote · full context-management configuration· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.3% |
| 5 | GPT-5 mini· BrowseComp-Plus data | 80% |
| 6 | Claude Opus 4.5Qwen3-Embedding-8B index · tool-result clearing + memory/context reset · Sonnet 4.5 grader· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 72.9% |
| 7 | GPT-6 Luna· BrowseComp-Plus data | 71.9% |
| 8 | Claude Opus 4.5launch-page footnote · basic fetch configuration· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 70.5% |
| 9 | Tongyi DeepResearch 30B A3B· BrowseComp-Plus data | 68.1% |
| 10 | Claude Opus 4.5Qwen3-Embedding-8B index · tool-result clearing · no fetch · Sonnet 4.5 grader· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 67.6% |
| 11 | Claude Sonnet 4.5Qwen3-Embedding-8B index · tool-result clearing + memory/context reset · Sonnet 4.5 grader· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 67.2% |
| 12 | OpenAI o3· BrowseComp-Plus data | 65.9% |
| 13 | GLM 4.7· BrowseComp-Plus data | 64.7% |
| 14 | Claude Sonnet 4.5Qwen3-Embedding-8B index · tool-result clearing · no fetch · Sonnet 4.5 grader· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 60.4% |
| 15 | Mercury 3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 55.2% |
| 16 | Claude Haiku 4.5Qwen3-Embedding-8B index · tool-result clearing + memory/context reset · Sonnet 4.5 grader· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 54.7% |
| 17 | Claude Haiku 4.5Qwen3-Embedding-8B index · tool-result clearing · no fetch · Sonnet 4.5 grader· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 52.5% |
| 18 | GLM 4.6· BrowseComp-Plus data | 48.2% |
| 19 | Gemini 2.5 Flash· BrowseComp-Plus data | 39.9% |
| 20 | GPT 4.1· BrowseComp-Plus data | 36.4% |
| 21 | kimi-k2-0711-preview· BrowseComp-Plus data | 35.4% |
| 22 | Gemini 2.5 Pro· BrowseComp-Plus data | 29.5% |
| 23 | DeepSeek R1· BrowseComp-Plus data | 16.4% |
| 24 | Qwen3 32B· BrowseComp-Plus data | 10.7% |
Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats