Agentic Coding on SWE-Bench Verified (Pass@1)
79.6Pass@1Claude Sonnet 4.6
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude Sonnet 4.6# Total Params=-, # Active Params=-2026.05 | 79.6 | |
| DeepSeek-V4 Flash# Total Params=284B, # Active Params=13B2026.05 | 79 | |
| Claude Sonnet-4.52026.03 | 77.2 | |
| Gemini 3-pro2026.03 | 76.2 | |
| Qwen3.5# Total Params=397B, # Active Params=17B2026.05 | 76.2 | |
| GPT-5 High2026.03 | 74.9 | |
| LAGUNA M.1# Total Params=225B, # Active Params=23B2026.05 | 74.6 | |
| GLM-4.7# Total Params=355B, # Active Params=32B2026.05 | 73.8 | |
| Qwen3.6# Total Params=35B, # Active Params=3B2026.05 | 73.4 | |
| Claude Haiku 4.52026.05 | 73.3 | |
| Seed1.82026.03 | 72.9 | |
| Devstral 2# Total Params=123B, # Active Params=123B2026.05 | 72.2 | |
| LAGUNA XS.2# Total Params=33.4B, # Active Params=3B2026.05 | 69.9 | |
| Qwen3.5# Total Params=35B, # Active Params=3B2026.05 | 69.2 | |
| Devstral Small 2# Total Params=24B, # Active Params=24B2026.05 | 68 | |
| Gemini 2.5-pro2026.03 | 59.6 | |
| Gemma 4# Total Params=31B, # Active Params=31B2026.05 | 52 |