Pyyan / Compare / YOLOv12 vs RF-DETR vs RTMDet vs Grounding DINO
YOLOv12 vs RF-DETR vs RTMDet vs Grounding DINO
4 of 5Object Detection · verified 13 Aug 2026
| Specification | YOLOv12 | RF-DETR | RTMDet | Grounding DINO |
|---|---|---|---|---|
| Summary | Attention-centric YOLO, still widely deployed. | First real-time detector past 60 mAP on COCO. | Pure throughput, and an MIT licence. | Detects whatever you describe in words. |
| COCO mAP | ~55 | 60.5 | ~52 | ~52 zero-shot |
| Speed | Very high | Real time | 300+ | Moderate |
| Licence | AGPL-3.0 | Apache 2.0 | MIT | Apache 2.0 |
| Family | Single stage | Transformer, set prediction | Single stage | Open vocabulary |
| Open vocabulary | No | No | No | Yes |
| Released | 2025 | 2025 | 2022 | 2023 |
| Category | Object Detection | Object Detection | Object Detection | Object Detection |
| Official | Ultralytics ↗ | Roboflow ↗ | OpenMMLab ↗ | IDEA Research ↗ |
Highlighted rows are where these differ.
YOLOv12
- Attention added to the classic single-stage design
- Same licensing consideration as YOLO26
Best for teams already on the YOLO toolchain.
RF-DETR
- Leads the RF100-VL domain-transfer benchmark
- DINOv2 backbone, markedly better on occluded objects
- Apache 2.0, which YOLO is not
Best for most custom detection tasks.
RTMDet
- Wins on raw speed where accuracy is sufficient
- MIT, so no derivative-work obligation
Best for high frame-rate video.
Grounding DINO
- Text prompt in, boxes out
- Fine-tuning still beats it on specialised objects
Best for classes you have no data for.