SWE-Marathon Leaderboard

Benchmark Model Rank Results
code-generation-on-swe-marathonClaude Sonnet 5.5 (max) (Claude Code)#1Resolution Rate: 51.25
code-generation-on-swe-marathonClaude Opus 5 (max) (Claude Code)#2Resolution Rate: 50.0
code-generation-on-swe-marathonKimi K3 (Claude Code)#3Resolution Rate: 48.13
code-generation-on-swe-marathonClaude Opus 5.5 (max) (Claude Code)#4Resolution Rate: 46.25
code-generation-on-swe-marathonGPT-5.6 Sol (max) (Codex)#5Resolution Rate: 42.5
code-generation-on-swe-marathonGLM-5.3 (max) (Claude Code)#6Resolution Rate: 42.5
code-generation-on-swe-marathonGPT-6 Astra (max) (Codex)#7Resolution Rate: 42.5
code-generation-on-swe-marathonGLM-5.3-Flash (max) (Claude Code)#8Resolution Rate: 38.75
code-generation-on-swe-marathonGPT-5.6 Terra (max) (Codex)#9Resolution Rate: 32.5
code-generation-on-swe-marathonGPT-5.6 Luna (max) (Codex)#10Resolution Rate: 24.38
code-generation-on-swe-marathonGemini 3.7 Flash (OpenCode)#11Resolution Rate: 22.5
code-generation-on-swe-marathonGLM-5.2 (Claude Code)#12Resolution Rate: 19.38
code-generation-on-swe-marathonDeepSeek V4 Pro (MiniSWE-Agent)#13Resolution Rate: 10.63
code-generation-on-swe-marathonClaude Fable 5.1 (max) (Claude Code)–Resolution Rate: 45.62
code-generation-on-swe-marathonClaude Haiku 5.5 (max) (Claude Code)–Resolution Rate: 50.0
code-generation-on-swe-marathonClaude Opus 4.8 (Claude Code)–Resolution Rate: 48.75
code-generation-on-swe-marathonClaude Sonnet 5 (max) (Claude Code)–Resolution Rate: 30.0
code-generation-on-swe-marathonGPT-6 Luna (max) (Codex)–Resolution Rate: 28.12
code-generation-on-swe-marathonGPT-6 Sol (max) (Codex)–Resolution Rate: 43.12
code-generation-on-swe-marathonGPT-6.1 Sol (max) (Codex)–Resolution Rate: 42.5
code-generation-on-swe-marathonGrok 4.5 (Grok Build)–Resolution Rate: 29.38
code-generation-on-swe-marathonGrok 4.6 (high) (Grok Build)–Resolution Rate: 31.88
code-generation-on-swe-marathonGrok 4.7 (high) (Grok Build)–Resolution Rate: 45.62
code-generation-on-swe-marathonMuse Spark 1.3 (xhigh) (Muse Code)–Resolution Rate: 30.62
code-generation-on-swe-marathonMuse Wonder (max) (Muse Code)–Resolution Rate: 52.5
code-generation-on-swe-marathonMuse Wonder Full (max) (Muse Code)–Resolution Rate: 53.12