Software Engineering Issue Resolution on SWE-Bench Lite (Pass@1)
53.7Pass@1Context Weaver
Evaluation Results
| Method | Links | |
|---|---|---|
| Context WeaverModel=Claude Sonnet 4, Setting=Unified2026.04 | 53.7 | |
| LLM SummarizationModel=Claude Sonnet 4, Setting=Unified2026.04 | 53 | |
| Sliding WindowModel=Claude Sonnet 4, Setting=Unified, window size=52026.04 | 52.3 | |
| Context WeaverModel=GPT-5, Setting=Hybrid2026.04 | 51.3 | |
| Sliding WindowModel=GPT-5, Setting=Unified, window size=52026.04 | 48.3 | |
| Context WeaverModel=Gemini 3 Flash, Setting=Unified2026.04 | 47 | |
| Sliding WindowModel=Gemini 3 Flash, Setting=Unified, window size=52026.04 | 46 | |
| LLM SummarizationModel=Gemini 3 Flash, Setting=Unified2026.04 | 43.3 | |
| Context WeaverModel=GPT-5, Setting=Unified2026.04 | 42 | |
| LLM SummarizationModel=GPT-5, Setting=Unified2026.04 | 32.3 |