Agentic benchmark
BrowseComp leaderboard
Every model the catalog carries a published BrowseComp value for, ranked by that value.
CategoryAgentic
MeasureWeb search and evidence synthesis
TasksResearch questions requiring browsing
DifficultyHard web research
Published byBrowseComp
A benchmark for web-browsing agents that must search, inspect sources, gather evidence, and return the correct answer to research-oriented questions.
BrowseComp ranking
43 models with a published BrowseComp value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Web search and evidence synthesis |
|---|---|---|---|
| 1 |  Atria Dawn Preview | Shanghai Artificial Intelligence Laboratory | 92.5 |
| 2 |  GPT-5.6 Sol | OpenAI | 92.2 |
| 3 | GPT-6 Astra | OpenAI | 91.5 |
| 4 |  Kimi K3 | Moonshot AI | 91.2 |
| 5 |  Claude Opus 5 | Anthropic | 90.8 |
| 6 | GPT-5.5 Pro | OpenAI | 90.1 |
| 7 | GPT-5.4 Pro | OpenAI | 89.3 |
| 8 | Claude Mythos 5 | Anthropic | 88 |
| 9 | GPT-5.6 Terra | OpenAI | 87.5 |
| 10 | OAOrnith-1.5-397B | Ornith AI | 86.6 |
| 11 | Claude Sonnet 5 | Anthropic | 84.7 |
| 12 | GPT-5.5 | OpenAI | 84.4 |
| 13 | Claude Opus 4.8 | Anthropic | 84.3 |
| 14 | Claude Opus 4.6 | Anthropic | 83.7 |
| 15 |  MiniMax M3 | MiniMax | 83.52 |
| 16 |  DeepSeek V4 Pro 0813 | DeepSeek | 83.4 |
| 17 | DSdots3-note Preview | Dots Studio | 83.3 |
| 17 | GPT-5.6 Luna | OpenAI | 83.3 |
| 19 | Kimi K2.6 | Moonshot AI | 83.2 |
| 20 | GPT-5.4 | OpenAI | 82.7 |
| 21 | Claude Opus 4.7 (Adaptive) | Anthropic | 79.3 |
| 22 | TMInkling-Small | Thinking Machines Lab | 77.4 |
| 23 | TMInkling | Thinking Machines Lab | 77.1 |
| 24 |  Step 3.7 Flash | StepFun | 75.82 |
| 25 | Agents-A1 | InternScience | 75.51 |
| 26 | DeepSeek V4 Flash 0731 | DeepSeek | 73.2 |
| 27 |  Ling 3.0 Flash | InclusionAI | 72.2 |
| 28 |  GLM-5.1 | Z.AI | 68 |
| 29 | OAOrnith-1.5-35B-A3B | Ornith AI | 67.6 |
| 30 | Agents-A1-4B | InternScience | 66.8 |
| 31 | GPT-5.2 | OpenAI | 65.8 |
| 32 |  Qwen3.5-122B-A10B | Alibaba | 63.8 |
| 33 | Qwen3.5 397B | Alibaba | 62 |
| 34 | Qwen3.5-27B | Alibaba | 61 |
| 34 | Qwen3.5-35B-A3B | Alibaba | 61 |
| 36 | Kimi K2.5 | Moonshot AI | 60.6 |
| 36 | Kimi K2.5 (Reasoning) | Moonshot AI | 60.6 |
| 38 | OAOrnith-1.5-9B | Ornith AI | 56.4 |
| 39 | GLM-4.7 | Z.AI | 52 |
| 40 |  Solar Pro 4 | Upstage | 49.2 |
| 41 |  LongCat-Flash-Lite-Sparse | Meituan | 48.62 |
| 42 |  Nemotron 3 Ultra | NVIDIA | 44.4 |
| 43 | Nemotron 3.5 Lightning 30B A3B NVFP4 | NVIDIA | 36.81 |
Evidence key: Observed
Rows are ordered by the value BrowseComp published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.