Datacurve AI

DeepSWE

A long-horizon software engineering benchmark from Datacurve for measuring frontier coding agents on original tasks drawn from active open-source repositories.

In index
Agentic
Best in agentic
Source ↗

Models59

Top result79.0%Muse Wonder Full

Top-3 spread1.1 pts

Year2026

Result and price

48 of 59 results

Results

59 results

#ModelResult

01Muse Wonder Full79.0%

02Muse Wonder78.8%

03Gemini 4 Argon77.9%

04DeepSeek V4.1 Flash 101777.0%

05Kimi K3.1 Pro75.5%

06Muse Spark 1.375.4%

07Ember-175.2%

08Hy474.9%

09Claude Fable 5.274.3%

10Claude Opus 5.574.2%

11GPT-6 Astra74.1%

12Gemini 3.8 Flash73.8%

13MiniMax M3.1 Flash73.7%

14SWE-273.0%

15GPT-5.6 Sol72.7%

16Particle-172.6%

17Inkling 2 Small72.1%

18GPT-6.1 Sol71.9%

18MiMo-V2.6-Pro71.9%

20ERNIE 5.5 Flash Preview71.2%

21Claude Sonnet 5.571.0%

21Grok 4.771.0%

23GPT-5.6 Terra69.6%

24Claude Opus 568.8%

24GPT-6 Sol68.8%

26MiMo-V2.6-Flash67.9%

27Step 5 Preview67.7%

28Kimi K367.5%

29Claude Fable 5.167.4%

30GPT-5.6 Luna67.2%

31GLM-5.366.9%

32GPT-6 Luna66.6%

33Grok 4.665.9%

34Gemini 3.7 Flash65.3%

35LongCat-2.564.6%

36Hy4 preview64.3%

37GLM-5.3-Flash63.4%

38DeepSeek V4 Pro 081362.7%

39Ultramarine 1 Flash61.5%

40SWE-2 Flash60.1%

41Muse Spark Open60.0%

42Ling 3.1 Flash59.7%

43Muse Spark 1.259.3%

44Qwen3.8-Flash-Next58.7%

45Qwen3.8-Omni-Flash57.8%

46Qwen3.8 Max56.6%

47Ornith-1.5-397B56.0%

48DeepSeek V4 Flash 073154.4%

49Muse Spark 1.153.3%

50Grok 4.553.0%

51Gemini 3.6 Flash49.0%

52Mercury 346.3%

53Ripple-65B Preview44.3%

54Beam 1.0 501B-A23B43.3%

55Beam 1.0 501B-A23B FP843.0%

56Beam 1.0 501B-A23B NVFP442.7%

57Qwen3.8-27B42.2%

58Laguna S 2.140.4%

59Ornith-1.5-35B-A3B22.0%