Tencent releases WeVisOne-VL 4B vision-language model on Hugging Face
Tencent’s WeChat Vision team has published WeVisOne-VL-4B, a unified vision‑language model that combines object‑level tasks such as open‑vocabulary detection, referring localization, counting, pointing and OCR with pixel‑level tasks like referring segmentation and promptable alpha matting. The model builds on Qwen3‑VL and uses structured next‑token prediction together with a lightweight 9.6…
Key points
- Tencent's WeChat Vision released WeVisOne-VL-4B, a unified vision‑language model built on Qwen3‑VL.
- Model supports detection, localization, counting, OCR and matting via structured next‑token prediction and a 9.6 M‑parameter pixel decoder.
- WeVisOne-VL-4B is available on Hugging Face under Apache 2.0 for research use, with code and weights publicly released.
The release includes an inference package on GitHub and a Hugging Face repository (tencent/WeVisOne-VL-4B). Users can load the model in a few lines of Python, specify a task (e.g., "matting"), and receive structured outputs such as masks and instance tokens. The code is under the Apache License 2.0, with third‑party components retaining their original licenses. The authors note that the model is intended for research in visual grounding and should not be used for autonomous high‑impact decisions or personal identification, and that fine alpha boundaries may be sensitive to image resolution and prompt phrasing.
tencent/WeVisOne-VL-4B · Hugging Face
huggingface.co · 23 October 2026
Loading the full article…
This text was published by huggingface.co. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Reddit discussionreddit.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Anthropic's Claude Opus 5.5 tops Coding Agent Index with 61.7 score · 1 src
- Nvidia and University of Washington unveil open-source protein design model RFD4-Proteina · 1 src
- Gradium open-sources Phonon 31M on-device TTS model under Apache 2.0 · 1 src
- Cohere releases Rerank 5 with passage-level context focus and calibrated scores · 1 src
- OpenAI says its AI agents weakened internal misalignment monitoring system · 31 src
Comments
via GitHub Discussions