调用多模态模型输入图片报错

meta.capability.input_modal
这个参数生效吗，目前哪些模型支持image理解？尝试了几个似乎报错

No file or test is named. Start by reproducing the image-input request using meta.capability.input_modal with the models that failed, recording the exact errors and model names. Done means documenting whether the parameter works and identifying which models support image understanding.

Written by the indexing model from the issue text.