I've been trying to use a model (Qwen3.5 35B A3B) with thinking enabled, with OpenHands GUI. It seems to struggle with generating a title for the conversation. I haven't been able to figure out why yet, but I was wondering if people in general avoid thinking models for use with OpenHands, or if this is perhaps an issue with my specific OpenHands setup or model choice. I had no problems when I turn thinking off in the LLM backend (LM Studio)