List your company

Software Engineering Problem Solving on SWE-bench Verified (Pass@K & Cost)

78Average Success Rate @1DELM
48.15255.90163.6571.399Jun 9, 2026

Evaluation results

MethodLinks
DELMBase Model=Claude Opus 4.62026.067880.782.50.63
mini-SWE-agentBase Model=Claude Opus 4.62026.0676.979.881.70.61
Claude CodeBase Model=Claude Opus 4.62026.0676.179.581.30.69
AOrchestra-ParallelBase Model=Claude Opus 4.62026.0675.278.179.70.73
AOrchestraBase Model=Claude Opus 4.62026.0674.778.380.10.7
DELMBase Model=Gemini 3 Flash2026.0665.772.977.40.12
AOrchestra-ParallelBase Model=Gemini 3 Flash2026.0656.463.271.80.25
AOrchestraBase Model=Gemini 3 Flash2026.0655.264.573.20.24
mini-SWE-agentBase Model=Gemini 3 Flash2026.0654.765.675.10.26
Claude CodeBase Model=Gemini 3 Flash2026.0649.357.166.3—