NVIDIA · Visual Understanding
Locate Anything (NVIDIA Research)
Unified vision-language grounding/detection model (Moon-ViT encoder + Qwen2.5 decoder) using Parallel Box Decoding to predict full bounding boxes in one step instead of token-by-token. Trained on 138M queries / 785M boxes — 12.7 boxes/sec (>10x faster than Qwen3-VL), +3.8% mean F1 on LVIS, SOTA on ScreenSpot-Pro (60.3) and M6Doc (70.1).
Open the source ↗Similar Vision & Image releases
- WeVisOne-VL
WeChat Vision, Tencent · Vision-Language Model · Week 44, 2026
- Nemotron 3.5 Super VL
NVIDIA · Open Multimodal MoE · Week 43, 2026
- Vision Demos
jeremyipark · Sports Vision Demos · Week 41, 2026
- PixelUMM
NVIDIA and University of Waterloo · Encoder-Free Unified Model · Week 41, 2026
- Ovis-Embedding
ATH-MaaS · Ovis Embedding Models · Week 40, 2026
- RADAR
RADAR-Generalist team, published in Science · Abdominal CT Vision Model · Week 40, 2026
More from NVIDIA
- NVIDIA Nemotron 3.5 Transcribe 30B-A4B
NVIDIA · Open multilingual ASR · Week 45, 2026
- SimReady Assets
NVIDIA · Simulation asset dataset · Week 45, 2026
- Nemotron 3.5 Super VL
NVIDIA · Open Multimodal MoE · Week 43, 2026
- SkillSpector v2.11.0
NVIDIA · Agent Skill Scanner · Week 36, 2026
- NVIDIA AVO on ARC-AGI-3
NVIDIA · Agent Orchestration · Week 34, 2026
- NVIDIA Nsight AI
NVIDIA · AI Infrastructure · Week 33, 2026
Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.