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.

In index
Agentic
Best in agentic

Models63

Top result92.5%Atria Dawn Preview

Top-3 spread0.4 pts

Year—

Result and price

40 of 63 results

Results

63 results

#ModelResult

01Atria Dawn Preview92.5%

02GPT-5.6 Sol92.2%

03Kimi K3.1 Pro92.1%

04LongCat-2.591.9%

05Ling 3.1 Flash91.7%

06GPT-6 Astra91.5%

07Kimi K391.2%

08Claude Opus 590.8%

09Inkling 2 Small90.2%

10GPT-5.5 Pro90.1%

11GPT-5.4 Pro89.3%

12Step 5 Preview88.7%

13Claude Mythos 588.0%

14GPT-5.6 Terra87.5%

15ERNIE 5.5 Flash Preview87.4%

16Ornith-1.5-397B86.6%

17Claude Sonnet 584.7%

18Mistral Large 484.5%

19GPT-5.584.4%

20Claude Opus 4.884.3%

21Claude Opus 4.683.7%

22MiniMax M383.5%

23DeepSeek V4 Pro 081383.4%

24GPT-5.6 Luna83.3%

24dots3-note Preview83.3%

26Kimi K2.683.2%

27GPT-5.482.7%

28Sage-A1.581.7%

29Ling 3.0 Flash Sante80.2%

30Claude Opus 4.7 (Adaptive)79.3%

31Nemotron 3.5 Super VL 120B A12B78.0%

32Nemotron 3.5 Super VL 120B A12B NVFP477.5%

33Inkling-Small77.4%

34Inkling77.1%

35Beam 1.0 501B-A23B76.8%

36Beam 1.0 501B-A23B FP876.7%

37Step 3.7 Flash75.8%

38Beam 1.0 501B-A23B NVFP475.6%

39Agents-A175.5%

40DeepSeek V4 Flash 073173.2%

41Ling 3.0 Flash72.2%

42Nemotron 3.5 Bolt 74B-A9B BF1670.0%

43Qwen3.5 397B69.0%

44GLM-5.168.0%

45Ornith-1.5-35B-A3B67.6%

46Agents-A1-4B66.8%

46Nemotron 3.5 Bolt 74B-A9B66.8%

48GPT-5.265.8%

49Qwen3.5-122B-A10B63.8%

50GLM-562.0%

51Qwen3.5-27B61.0%

51Qwen3.5-35B-A3B61.0%

53Kimi K2.5 (Reasoning)60.6%

53Kimi K2.560.6%

55Ornith-1.5-9B56.4%

56Spark-X2.6-4B53.1%

57GLM-4.752.0%

58Sarvam 105B49.5%

59Solar Pro 449.2%

60LongCat-Flash-Lite-Sparse48.6%

61Nemotron 3 Ultra44.4%

62Nemotron 3.5 Lightning 30B A3B NVFP436.8%

63Sarvam 30B35.5%