Tencent Releases WeVisOne-VL Unified Vision Language Model
Original title: Tencent WeVisOne-VL Unifies Detection, OCR, Segmentation and Matting
Tencent's WeVisOne-VL handles detection, OCR, segmentation and matting in one open model for 2B and 4B sizes.
WeVisOne-VL extends Qwen3-VL with a 9.6M-parameter pixel decoder that outputs boxes, points, masks or mattes. It rejects absent targets as None. Trained on 100M images with 3.6B boxes across three SFT stages and GRPO. The model outperforms baselines on 30+ benchmarks including LVIS and TotalText.
- 2B and 4B parameter versions available on Hugging Face.
- 71.3 F1 on Dense200 dense detection.
- 52.0 F1 on TotalText OCR.
- 9.6M-parameter decoder enables native masks.
- Vendors
- Tencent
- Products
- WeVisOne-VL
- AI models
- WeVisOne-VL-2BWeVisOne-VL-4B
From the original article
7 min read at marktechpost.comTencent’s WeChat Vision team released WeVisOne-VL, an Apache 2. 0 vision language model in 2B and 4B sizes. It handles detection, OCR, counting, segmentation and alpha matting in one model, with a 9.
By Michal Sutter · marktechpost.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.