Qwen 3.8 mimics GPT-5.5 Pro's thinking, not Opus 4.8

Original: Qwen 3.8 follows GPT-5.5 Pro reasoning prefills

Source: gist.github.com ↗