There is now an “Inference” option, but even if “Inference” is checked, the inference (deep thinking) mode for the DeepSeek—v3.1 and DeepSeek—v3.2 models in SiliconFlow cannot be turned on. After testing, models with inference modes by default in SiliconFlow (such as DeepSeek-R1, Kimi-K2-Thinking, etc.) will automatically execute the inference mode even if “inference” is not checked in Chatbox; for models with inference mode in SiliconFlow but with inference mode turned off by default (such as DeepSeek-v3.2), Chatbox will not execute inference mode even if “inference” is checked in chatbox. Whether to check “Inference” seems to have the same result.