Software Engineering on SWE-bench
82.4Resolve RateAgentOrchestra
Evaluation Results
| Method | Links | |
|---|---|---|
| AgentOrchestra2026.05 | 82.4 | |
| Uno-Orchestra2026.05 | 81.8 | |
| Qwen3.5-122B-A10BFramework/Scaffold=OpenCode2026.04 | 67.4 | |
| Qwen3.5-122B-A10BFramework/Scaffold=OpenHands2026.04 | 66.4 | |
| AOrchestra2026.05 | 61.7 | |
| Qwen3.5-122B-A10BFramework/Scaffold=Codex2026.04 | 61.2 | |
| Nemotron 3 SuperFramework/Scaffold=OpenHands2026.04 | 60.47 | |
| Nemotron 3 SuperFramework/Scaffold=OpenCode2026.04 | 59.2 | |
| Devstral Small 2Model size=24B, Model architecture=Dense2026.04 | 56.4 | |
| DevinContext=Best Published Baseline2026.03 | 55 | |
| Nemotron 3 SuperFramework/Scaffold=Codex2026.04 | 53.73 | |
| GPT-OSS-120BFramework/Scaffold=OpenHands2026.04 | 41.9 | |
| gpt-oss-20bModel size=20B, Model architecture=MoE2026.04 | 26 | |
| xRouter2026.05 | 24.8 | |
| ColdStart-LLM2026.05 | 14.6 | |
| ARYAParameters=02026.03 | 0 |