Agentic Coding on SWE-bench Multilingual (pass@1)
73.3pass@1DeepSeek-V4 Flash
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepSeek-V4 Flash# Total Params=284B, # Active Params=13B2026.05 | 73.3 | |
| Qwen3.5# Total Params=397B, # Active Params=17B2026.05 | 69.3 | |
| GLM-4.7# Total Params=355B, # Active Params=32B2026.05 | 66.7 | |
| LAGUNA M.1# Total Params=225B, # Active Params=23B2026.05 | 63.1 | |
| Devstral 2# Total Params=123B, # Active Params=123B2026.05 | 61.3 |