Benchmark
BrowseComp
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.
- Models measured
- 84
- Scores recorded
- 203
- Sources
- 4
- IQ weight
- 2.43
Score distribution
- 66.4%
- Mean
- 74.7%
- Median
- 21.8%
- Std dev
- 35.2%
- 10th percentile
- 90.1%
- 90th percentile
- 0.6%
- Min
This benchmark has a weight of 2.43 for chat models, measured over 84 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 |
|---|---|---|
| 151 | Qwen3 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 53.9% |
| 152 | DeepSeek V4 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 53.5% |
| 153 | DeepSeek V4 FlashDeepSeek V4 official model card · Think High· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 53.5% |
| 154 | DeepSeek V3.2Thinking mode · temperature 1.0 · 128K context · commercial search API · baseline · Section 4.4· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 53.4% |
| 155 | Claude Opus 5OpenRouter harness, anthropic/claude-opus-5, exa, server-tool, 5, high· OpenRouter benchmark exports | 52.5% |
| 156 | Step 3.5 FlashStep 3.5 Flash official headline table · vanilla· Reported by StepFun· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 51.6% |
| 157 | o4 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 51.5% |
| 158 | DeepSeek Reasoner· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 51.4% |
| 159 | DeepSeek V3.2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 51.4% |
| 160 | DeepSeek V3.2Thinking mode · temperature 1.0 · 128K context · commercial search API · without context management · Table 2· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 51.4% |
| 161 | OpenAI o3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 49.7% |
| 162 | Sarvam 105B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 49.5% |
| 163 | Step 3.5 FlashStep 3.5 technical report · Table 17 · 200-item subset · plain context · 86 real steps· Reported by StepFun· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 49.5% |
| 164 | Solar Pro 4· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 49.2% |
| 165 | Mistral Medium 3.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 48.6% |
| 166 | GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna, exa, server-tool, 5, xhigh· OpenRouter benchmark exports | 48% |
| 167 | GPT-5.6 SolOpenRouter harness, openai/gpt-5.6-sol, exa, server-tool, 1, high· OpenRouter benchmark exports | 46.7% |
| 168 | GPT-5.6 SolOpenRouter harness, openai/gpt-5.6-sol, perplexity, server-tool, 1, high· OpenRouter benchmark exports | 46.3% |
| 169 | MiMo V2 FlashMiMo-V2-Flash final evaluation · standard· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 45.4% |
| 170 | GLM 4.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 45.1% |
| 171 | Grok 4 Fast· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 44.9% |
| 172 | Nemotron 3 Ultra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 44.4% |
| 173 | Nemotron 3 UltraNeMo Evaluator SDK · official table· Reported by NVIDIA· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 44.4% |
| 174 | MiniMax M2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.Values from 2 sources
| 44% |
| 175 | GPT-5.6 SolOpenRouter harness, openai/gpt-5.6-sol, parallel, server-tool, 1, high· OpenRouter benchmark exports | 43.7% |
| 176 | GLM 4.7 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 42.8% |
| 177 | GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna, parallel, server-tool, 5, xhigh· OpenRouter benchmark exports | 42.5% |
| 178 | GPT-5.6 SolOpenRouter harness, openai/gpt-5.6-sol, native, server-tool, 1, high· OpenRouter benchmark exports | 41.8% |
| 179 | DeepSeek V3.2 Exp· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 40.1% |
| 180 | Nemotron 3.5 Lightning 30B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 37% |
| 181 | Nemotron 3.5 Lightning 30B A3Bofficial release table· Reported by NVIDIA· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 37% |
| 182 | Claude Opus 5OpenRouter harness, anthropic/claude-opus-5, perplexity, server-tool, 1, high· OpenRouter benchmark exports | 35.8% |
| 183 | Sarvam-30B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 35.5% |
| 184 | DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-0731, parallel, server-tool, 5, high· OpenRouter benchmark exports | 35.2% |
| 185 | GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna, perplexity, server-tool, 1, xhigh· OpenRouter benchmark exports | 33.7% |
| 186 | GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna, exa, server-tool, 1, xhigh· OpenRouter benchmark exports | 33% |
| 187 | Nemotron 3 Super 120B A12B· Steel.dev Agent Leaderboards dataValues from 2 sources
| 31.3% |
| 188 | Nemotron 3 Super 120B A12BNeMo Evaluator / mostly NeMo Skills · with Search · internal SerpAPI implementation· Reported by NVIDIA· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 31.3% |
| 189 | Claude Opus 5OpenRouter harness, anthropic/claude-opus-5, parallel, server-tool, 1, high· OpenRouter benchmark exports | 30.8% |
| 190 | DeepSeek V3.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 30% |
| 191 | Claude Opus 5OpenRouter harness, anthropic/claude-opus-5, exa, server-tool, 1, high· OpenRouter benchmark exports | 29.4% |
| 192 | GLM 4.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 26.4% |
| 193 | GPT-5.6 LunaOpenRouter harness, openai/gpt-5.6-luna, parallel, server-tool, 1, xhigh· OpenRouter benchmark exports | 23.7% |
| 194 | GLM 4.5 Air· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 21.3% |
| 195 | DeepSeek V4 Flash 0731OpenRouter harness, deepseek/deepseek-v4-flash-0731, parallel, server-tool, 1, high· OpenRouter benchmark exports | 20% |
| 196 | Claude Opus 5OpenRouter harness, anthropic/claude-opus-5, exa, plugin, medium· OpenRouter benchmark exports | 12% |
| 197 | Claude Opus 5OpenRouter harness, anthropic/claude-opus-5, perplexity, plugin, medium· OpenRouter benchmark exports | 11% |
| 198 | o1· Steel.dev Agent Leaderboards data | 9.9% |
| 199 | DeepSeek R1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 8.9% |
| 200 | Sonar Deep Research· Steel.dev Agent Leaderboards data | 6% |
Showing 151–200 of 203 results.
Data credits: Steel.dev Agent Leaderboards, MIT (licence) · Data by LLM Stats · 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. · OpenRouter public benchmark leaderboard and search-lane exports, CC BY 4.0. Fractional source scores converted to percentages; immutable dated snapshots. (licence)