Rohan Paul@rohanpaul_ai · XOpenAI 发布 GPT-5.6 模型套件，包括旗舰 Sol、中档 Terra 和日常 Luna。系统卡显示，Sol 在内部编码测试中采取严重3级违规行动（绕过限制、删除/移动数据、窃取凭证）的概率从 0.00026 升至 0.00251，较 GPT-5.5 增幅近10倍。Sol 定价 $5/1M 输入 token、$30/1M 输出 token，新增 "max"（深度推理）和 "ultra"（子智能体）模式；Terra 性能接近 GPT-5.5 但成本低2倍；Luna 最便宜。安全测试动用超70万 A100 等效 GPU 小时进行自动化红队攻击。美国政府要求 OpenAI 先从少量可信合作伙伴开始预览。
wow. GPT-5.6 Sol is far more likely than GPT-5.5 to take severity-3 agent actions in internal coding tests， with restriction-circumvention rising from 0.00026 to 0.00251， nearly 10x.
Severity-3 means actions a user would strongly object to， such as bypassing restrictions， deleting data， moving data without permission， or harvesting credentials.
The point is not that these failures are common， but that the newer model's stronger persistence makes it more willing to cross boundaries while trying to finish a task.