InternVL
A vision-language model family for self-hosted visual chat and image retrieval, with downloadable weights and Hugging Face Transformers support.
InternVL is a family of vision-language models for developers and researchers building AI applications that interpret images and answer questions about them. Downloadable models let teams host inference themselves.
- Visual chat: Build conversations around images, including questions about documents, charts and visual mathematics.
- Classification and retrieval: Use vision-language models for image and video classification, or find images and videos through text queries.
- Visual recognition: InternViT vision components support image-level and pixel-level recognition for computer vision applications beyond chat.
- Model choices: Select smaller or larger variants with language backbones including Qwen2.5 and InternLM, paired with InternViT vision encoders.
Checkpoints are available through Hugging Face and ModelScope. The project provides its own checkpoint format and a format compatible with Hugging Face Transformers. Its Python code is open source under the MIT license.
InternVL also has a hosted chat demo and an API. Those services run remotely; the downloadable checkpoints support deployment on your own infrastructure.
Logotyper
