{"version":1,"type":"story","url":"https://digestai.news/story/tencent-releases-wevisone-vl-a-2b-4b-open-vision-model-for-detection-o","json":"https://digestai.news/story/tencent-releases-wevisone-vl-a-2b-4b-open-vision-model-for-detection-o.json","markdown":"https://digestai.news/story/tencent-releases-wevisone-vl-a-2b-4b-open-vision-model-for-detection-o.md","slug":"tencent-releases-wevisone-vl-a-2b-4b-open-vision-model-for-detection-o","headline":"Tencent releases WeVisOne-VL, a 2B/4B open vision model for detection, OCR and matting","summary":"Tencent’s WeChat Vision team has open-sourced **WeVisOne-VL**, a vision-language model in 2B and 4B parameter sizes. It combines detection, object counting, referring, OCR, segmentation, and alpha matting into one model, eliminating the need for separate tools like SAM for pixel-level tasks. The model uses a 9.6M-parameter decoder for outputs and can return *“None”* when an object is absent, reducing false positives in queries.\n\nTrained on a knowledge-graph-driven dataset of 100M+ images with 3.6B annotated boxes, WeVisOne-VL outperforms baselines like LocateAnything-3B on benchmarks such as Dense200 (71.3 F1 vs. 58.7) and PixMo-Count (89.6% exact match). The 4B model leads on most tasks, while the 2B version still excels in box-prompted matting. Weights are available under Apache 2.0 on Hugging Face, with hosted demos and a Python wrapper for inference. Fine-tuning code is pending.","keyPoints":["Combines detection, OCR, segmentation, and matting into one 2B/4B model, replacing multiple tools","Outperforms LocateAnything-3B on Dense200 (71.3 F1 vs. 58.7) and PixMo-Count (89.6% exact match)","Open-sourced under Apache 2.0 with hosted demos and Hugging Face weights"],"whyItMatters":"A unified model for perception tasks could reduce costs and complexity for developers building computer vision pipelines, especially in applications requiring segmentation or matting alongside detection.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":["Tencent","Tencent WeChat Vision"],"models":["WeVisOne-VL","Qwen3-VL","LocateAnything-3B","SAM3","Rex-Omni","PP-OCRv5"],"people":["Michal Sutter"]},"firstPublishedAt":"2026-11-04T04:39:27Z","updatedAt":"2026-11-04T04:39:27Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"MarkTechPost","title":"Tencent WeVisOne-VL Unifies Detection, OCR, Segmentation and Matting","url":"https://marktechpost.com/2026/11/03/tencent-wevisone-vl-vision-language-model","publishedAt":"2026-11-04T04:39:27Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":{"title":"Tencent Open Source Vision Model Launch","url":"https://digestai.news/thread/tencent-releases-wevisone-vl-4b-vision-language-model-on-hugging-face","storyCount":2},"cite":{"text":"Digest AI, \"Tencent releases WeVisOne-VL, a 2B/4B open vision model for detection, OCR and matting\", 4 November 2026, https://digestai.news/story/tencent-releases-wevisone-vl-a-2b-4b-open-vision-model-for-detection-o","publisher":"Digest AI","title":"Tencent releases WeVisOne-VL, a 2B/4B open vision model for detection, OCR and matting","datePublished":"2026-11-04T04:39:27Z","url":"https://digestai.news/story/tencent-releases-wevisone-vl-a-2b-4b-open-vision-model-for-detection-o"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}