InternVL

InternVL is an open MLLM family from OpenGVLab, offered in sizes from 1B to 78B so you can pick the right model for everything from lightweight deployments to demanding production workloads.

Thanks to native multimodal pre-training, InternVL excels at vision, reasoning, long-context understanding, and agent tasks—handling the mixed text-and-image scenarios that everyday AI applications actually face.

Beyond multimodal strengths, InternVL even outperforms base LLMs on pure text tasks, so you don't have to trade away language quality to gain vision capabilities.