Qwen 3.6 27B on DeepSWE
Signal
65
Hype
25
In three linesQwen 3.6 27B scores 1.79% on DeepSWE, ranked 18th above Haiku 4.5. 70-hour benchmark on RTX6000 Blackwell with reasoning enabled and 262k context window. Average 44k output tokens per task, performance on par with similar models despite verbosity reputation.Read source
Your take?
Summary generated by Claude — human-verified