DeepSWE v1.1

Long-horizon software engineering on real codebases.

113Public tasks

75.2%Highest score

The DeepSWE v1.1 leaderboard

113 original software engineering tasks spanning 91 repositories and 5 languages. Each task requires large, novel fixes not sourced from existing public commits.

Paper Data Code Sample task

Model

Score

Muse Wonder 1.0Max

75.2%

GPT-6 AstraMax

73.2%

Kimi K3Max

72.9%

Opus 5.5Max

72.3%

GPT-6.1 SolMax

72.3%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Pass@1

DeepSWE v1.1

The DeepSWE v1.1 leaderboard

The latest on frontier AI performance, straight to your inbox.