Tencent WeChat Vision team open-sources WeVisOne-VL
Understand this story
Tencent's WeChat Vision team has open-sourced the vision-language model WeVisOne-VL, offering 2B (2.44B parameters) and 4B (4.83B parameters) versions that use a single model to perform detection, referring, counting, pointing, OCR, segmentation, and alpha matting. Weights for both versions have been released on Hugging Face. The models can run in BF16 on a single GPU via Transformers or vLLM. Fine-tuning code has not yet been released and remains on the to-do list; the GGUF version cannot generate masks because the decoder requires native visual features.
AI overview based on source reports · Updated 22 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- MarkTechPostTencent WeVisOne-VL Unifies Detection, OCR, Segmentation and Matting
Tencent's WeChat Vision team released WeVisOne-VL, an open vision language model in 2B and 4B sizes that handles detection, referring, counting, pointing, OCR, segmentation and alpha matting in one model, with pixel outputs from a 9.6M-parameter decoder so no separate SAM is needed. Built on Qwen3-VL, it is trained with up to 8,192 tokens of context and runs on a single GPU via Transformers or vLLM in BF16. Tencent reports the 4B model scoring 71.3 F1 on Dense200 and 81.8 F1 on RefCOCOg test, while noting a 52.6% rejection rate on PR-Bench.
Activity for this story
- Current comparable activity
- 3
- Peak comparable activity
- 9Nov 4 15:00
- Change in the past 24 hours
- -51%
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.