List your company

Agentic Software Engineering on SWE-bench Multilingual (SWE-M)

50Success RateClaude-Opus-4.5
-211.52538.5Jun 16, 2026

Evaluation results

MethodLinks
Claude-Opus-4.52026.0650
Gemini-3-Pro2026.0642.7
DeepSeek-V3.22026.0637.4
Kimi-K2-Instruct-09052026.0633.5
Qwen3-Coder-480B-A35B-Instruct2026.0632.7
LoopCoder-v2R (number of refinement loops)=22026.0631
LoopCoder-v2R (number of refinement loops)=12026.0614
LoopCoder-v2R (number of refinement loops)=32026.0611
LoopCoder-v2R (number of refinement loops)=42026.069.3
DeepSeek-Coder-V2-Lite-Instruct2026.060
Qwen2.5-Coder-7B-Instruct2026.060
Seed-Coder-8B-Instruct2026.060
Qwen2.5-Coder-14B-Instruct2026.060
Qwen2.5-Coder-32B-Instruct2026.060