Hacker News · AI·17d agoQwen 3.8 follows GPT-5.5 Pro reasoning prefills#gpt-5.5-pro#llm#prefillsAI research