Multimodal in-context learning capabilities of VLLMs are rigorously evaluated through the VL-ICL Bench, highlighting strengths and weaknesses.