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 |
|---|---|---|
| 1 | GPT-5.6 SolAtria Dawn Preview official card / report Table 1 · published comparison run · model snapshots and reasoning tiers not specified unless stated · in-house search/visit/python harness · 500 steps · discard-all context strategy· Reported by Shanghai AI Laboratory· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 92.2% |
| 2 | GPT-5.6 SolUltra effort· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 92.2% |
| 3 | Kimi K3.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 92.1% |
| 4 | GPT 5.2Long Context · 128K · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 92% |
| 5 | LongCat 2.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 91.9% |
| 6 | GPT-6 Astra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 91.5% |
| 7 | GPT-6 AstraOpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 91.5% |
| 8 | Kimi K3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 91.2% |
| 9 | Kimi K3Atria Dawn Preview official card / report Table 1 · published comparison run · model snapshots and reasoning tiers not specified unless stated · in-house search/visit/python harness · 500 steps · discard-all context strategy· Reported by Shanghai AI Laboratory· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 91.2% |
| 10 | Kimi K3context compaction at 300K · 1M context · max effort · temp 1.0 · top_p 1.0· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 91.2% |
| 11 | Claude Opus 5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 90.8% |
| 12 | Claude Opus 5Atria Dawn Preview official card / report Table 1 · published comparison run · model snapshots and reasoning tiers not specified unless stated · in-house search/visit/python harness · 500 steps · discard-all context strategy· Reported by Shanghai AI Laboratory· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.8% |
| 13 | Claude Opus 5OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.8% |
| 14 | Claude Opus 5Opus 5 System Card capability summary · adaptive thinking · max effort unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.8% |
| 15 | GPT-5.6 Sol· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 90.4% |
| 16 | GPT-5.6 SolOpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.4% |
| 17 | GPT-5.6 SolOpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.4% |
| 18 | GPT-5.6 SolOpus 5 System Card capability summary · adaptive thinking · max effort unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.4% |
| 19 | Kimi K3no context management · full 1M context · max effort · temp 1.0 · top_p 1.0· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.4% |
| 20 | Inkling 2 Small· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 90.2% |
| 21 | GPT-5.5 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 90.1% |
| 22 | GPT-5.5 Proxhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 90.1% |
| 23 | GPT 5.2Long Context · 256K · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 89.8% |
| 24 | GPT-5.4 Pro· Steel.dev Agent Leaderboards data | 89.3% |
| 25 | GPT-5.4 ProOpenAI GPT-5.4 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 89.3% |
| 26 | Claude Opus 5OpenRouter harness, anthropic/claude-opus-5, perplexity, server-tool, 25, high· OpenRouter benchmark exports | 89% |
| 27 | Claude Opus 4.8multi-agent· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 88.5% |
| 28 | Claude Fable 5· Steel.dev Agent Leaderboards data | 88% |
| 29 | GPT-5.6 Terra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.5% |
| 30 | GPT-5.6 TerraOpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.5% |
| 31 | Claude Fable 5OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.4% |
| 32 | Claude Fable 5Opus 5 System Card capability summary · adaptive thinking · max effort unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 87.4% |
| 33 | ERNIE 5.5 Flash Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.4% |
| 34 | Claude Sonnet 5multi-agent · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 86.6% |
| 35 | Claude Opus 5OpenRouter harness, anthropic/claude-opus-5, parallel, server-tool, 25, high· OpenRouter benchmark exports | 86.3% |
| 36 | Kimi K2.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 86.3% |
| 37 | Kimi K2.6Kimi K2.6 official benchmark table · thinking · temp 1.0 · top_p 1.0 · 262,144 context · Agent Swarm· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 86.3% |
| 38 | Doubao Seed 2.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 86.2% |
| 39 | Gemini 3.1 Pro Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 85.9% |
| 40 | Gemini 3.1 Pro PreviewFable/Mythos 5 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.9% |
| 41 | Gemini 3.1 Pro PreviewOpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.9% |
| 42 | Gemini 3.1 Pro PreviewOpus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.9% |
| 43 | Gemini 3.1 Pro PreviewSearch + Python + Browse· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.9% |
| 44 | Doubao Seed 2.1 Turbo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.9% |
| 45 | Claude Sonnet 5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 84.7% |
| 46 | Claude Sonnet 5single agent · search/fetch/code · 10M token limit · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.7% |
| 47 | Mistral Large 4· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.5% |
| 48 | GPT 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 84.4% |
| 49 | GPT 5.5Fable/Mythos 5 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.4% |
| 50 | GPT 5.5OpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.4% |
Showing 1–50 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)