Score: 2

Jina-VLM: Small Multilingual Vision Language Model

Published: December 3, 2025 | arXiv ID: 2512.04032v2

By: Andreas Koukounas , Georgios Mastrapas , Florian Hönicke and more

Potential Business Impact:

Lets computers answer questions about pictures.

Business Areas:
Natural Language Processing Artificial Intelligence, Data and Analytics, Software

We present Jina-VLM, a 2.4B parameter vision-language model that achieves state-of-the-art multilingual visual question answering among open 2B-scale VLMs. The model couples a SigLIP2 vision encoder with a Qwen3 language backbone through an attention-pooling connector that enables token-efficient processing of arbitrary-resolution images. The model achieves leading results on standard VQA benchmarks and multilingual evaluations while preserving competitive text-only performance. Model weights and code are publicly released at https://huggingface.co/jinaai/jina-vlm .


Page Count
18 pages

Category
Computer Science:
Computation and Language