InternLM-XComposer2 is a vision-language model that excels in free-form text-image composition and comprehension. Based on InternLM2-7B, it significantly outperforms existing multimodal models and matches or surpasses GPT-4V and Gemini Pro in certain assessments. The model uses a Partial LoRA (PLoRA) approach, applying additional LoRA parameters only to image tokens to preserve pre-trained language knowledge while balancing precise vision understanding with text composition. It is trained on high-quality, diverse data for free-form text-image composition and multimodal understanding, enabling it to generate high-quality, integrated text-image content from various inputs. InternLM-XComposer2 demonstrates exceptional performance in various benchmarks, including MathVista, MMMU, AI2D, MME, MMBench, and others. It outperforms existing open-source models and performs on par with closed-source APIs. The model's capabilities include detailed perception, logical reasoning, and extensive knowledge integration, making it highly effective for multimodal understanding. InternLM-XComposer2 is publicly available at https://github.com/InternLM/InternLM-XComposer.InternLM-XComposer2 is a vision-language model that excels in free-form text-image composition and comprehension. Based on InternLM2-7B, it significantly outperforms existing multimodal models and matches or surpasses GPT-4V and Gemini Pro in certain assessments. The model uses a Partial LoRA (PLoRA) approach, applying additional LoRA parameters only to image tokens to preserve pre-trained language knowledge while balancing precise vision understanding with text composition. It is trained on high-quality, diverse data for free-form text-image composition and multimodal understanding, enabling it to generate high-quality, integrated text-image content from various inputs. InternLM-XComposer2 demonstrates exceptional performance in various benchmarks, including MathVista, MMMU, AI2D, MME, MMBench, and others. It outperforms existing open-source models and performs on par with closed-source APIs. The model's capabilities include detailed perception, logical reasoning, and extensive knowledge integration, making it highly effective for multimodal understanding. InternLM-XComposer2 is publicly available at https://github.com/InternLM/InternLM-XComposer.