Tencent’s WeChat Vision team released WeVisOne-VL, a vision-language model that handles object detection, text recognition (OCR), counting, segmentation and alpha matting. The model comes in 2-billion- and 4-billion-parameter sizes and includes a 9.6-million-parameter pixel decoder. It is licensed under Apache 2.0, which permits broad use subject to the license’s terms. The announcement does not give benchmark results or say where people can download or run the model.
Why it matters
Developers can consider one Apache 2.0 model for five image tasks, but without access details or benchmark results, its practical performance is unclear.
