Skip to content

BrowseComp

BrowseComp tests whether a browsing agent can find difficult, fact-seeking answers across the web. Its 1,266 problems use short answers that are easier to grade than an open-ended research report.

50 models

Higher is better.

  1. 1Atria Dawn PreviewShanghai Artificial Intelligence Laboratory92.5
  2. 2GPT-5.6 SolOpenAI92.2
  3. 3Ling 3.1 FlashInclusionAI91.7
  4. 4GPT-6 AstraOpenAI91.5
  5. 5Kimi K3Moonshot AI91.2
  6. 6Claude Opus 5Anthropic90.8
  7. 7GPT-5.5 ProOpenAI90.1
  8. 8GPT-5.4 ProOpenAI89.3
  9. 9Step 5 PreviewStepFun88.7
  10. 10Claude Mythos 5Anthropic88.0
  11. 11GPT-5.6 TerraOpenAI87.5
  12. 12Ornith-1.5-397BOrnith AI86.6
  13. 13Claude Sonnet 5Anthropic84.7
  14. 14GPT-5.5OpenAI84.4
  15. 15Claude Opus 4.8Anthropic84.3
  16. 16Claude Opus 4.6Anthropic83.7
  17. 17MiniMax M3MiniMax83.5
  18. 18DeepSeek V4 Pro 0813DeepSeek83.4
  19. 19GPT-5.6 LunaOpenAI83.3
  20. 20dots3-note PreviewDots Studio83.3
  21. 21Kimi K2.6Moonshot AI83.2
  22. 22GPT-5.4OpenAI82.7
  23. 23Claude Opus 4.7 (Adaptive)Anthropic79.3
  24. 24Nemotron 3.5 Super VL 120B A12BNVIDIA78.0
  25. 25Inkling-SmallThinking Machines Lab77.4
  26. 26InklingThinking Machines Lab77.1
  27. 27Beam 1.0 501B-A23BReflection AI76.8
  28. 28Step 3.7 FlashStepFun75.8
  29. 29Agents-A1InternScience75.5
  30. 30DeepSeek V4 Flash 0731DeepSeek73.2
  31. 31Ling 3.0 FlashInclusionAI72.2
  32. 32Qwen3.5 397BAlibaba69.0
  33. 33GLM-5.1Z.AI68.0
  34. 34Ornith-1.5-35B-A3BOrnith AI67.6
  35. 35Nemotron 3.5 Bolt 74B-A9BNVIDIA66.8
  36. 36GPT-5.2OpenAI65.8
  37. 37Qwen3.5-122B-A10BAlibaba63.8
  38. 38GLM-5Z.AI62.0
  39. 39Qwen3.5-27BAlibaba61.0
  40. 40Qwen3.5-35B-A3BAlibaba61.0
Higher is better. Showing the top 40 of 50. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Web search and evidence synthesis
Tasks
Research questions requiring browsing
Authors
OpenAI (2025)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.

All results

1Atria Dawn Preview Shanghai Artificial Intelligence Laboratory92.5
2GPT-5.6 Sol OpenAI92.2
3Ling 3.1 Flash InclusionAI91.7
4GPT-6 Astra OpenAI91.5
5Kimi K3 Moonshot AI91.2
6Claude Opus 5 Anthropic90.8
7GPT-5.5 Pro OpenAI90.1
8GPT-5.4 Pro OpenAI89.3
9Step 5 Preview StepFun88.7
10Claude Mythos 5 Anthropic88.0
11GPT-5.6 Terra OpenAI87.5
12Ornith-1.5-397B Ornith AI86.6
13Claude Sonnet 5 Anthropic84.7
14GPT-5.5 OpenAI84.4
15Claude Opus 4.8 Anthropic84.3
16Claude Opus 4.6 Anthropic83.7
17MiniMax M3 MiniMax83.5
18DeepSeek V4 Pro 0813 DeepSeek83.4
19GPT-5.6 Luna OpenAI83.3
20dots3-note Preview Dots Studio83.3
21Kimi K2.6 Moonshot AI83.2
22GPT-5.4 OpenAI82.7
23Claude Opus 4.7 (Adaptive) Anthropic79.3
24Nemotron 3.5 Super VL 120B A12B NVIDIA78.0
25Inkling-Small Thinking Machines Lab77.4
26Inkling Thinking Machines Lab77.1
27Beam 1.0 501B-A23B Reflection AI76.8
28Step 3.7 Flash StepFun75.8
29Agents-A1 InternScience75.5
30DeepSeek V4 Flash 0731 DeepSeek73.2
31Ling 3.0 Flash InclusionAI72.2
32Qwen3.5 397B Alibaba69.0
33GLM-5.1 Z.AI68.0
34Ornith-1.5-35B-A3B Ornith AI67.6
35Nemotron 3.5 Bolt 74B-A9B NVIDIA66.8
36GPT-5.2 OpenAI65.8
37Qwen3.5-122B-A10B Alibaba63.8
38GLM-5 Z.AI62.0
39Qwen3.5-27B Alibaba61.0
40Qwen3.5-35B-A3B Alibaba61.0
41Kimi K2.5 (Reasoning) Moonshot AI60.6
42Kimi K2.5 Moonshot AI60.6
43Ornith-1.5-9B Ornith AI56.4
44GLM-4.7 Z.AI52.0
45Sarvam 105B Sarvam49.5
46Solar Pro 4 Upstage49.2
47LongCat-Flash-Lite-Sparse Meituan48.6
48Nemotron 3 Ultra NVIDIA44.4
49Nemotron 3.5 Lightning 30B A3B NVFP4 NVIDIA36.8
50Sarvam 30B Sarvam35.5