public good

Cross-chip benchmarks

Every model in the reference set, run on every chip that supports it. Sortable. Reproducible. Updated weekly.

195 benchmark entries · 10 task families · 17 chips

Object detection · cross-chip leaderboard · sorted by latency p50

Object detection

#	Model	Chip	Quant	Latency p50	Throughput	Acc.	Power
1	NVIDIA-YOLO27-Orin	NNVIDIA Jetson Thor	MIXED	3.92 ms	255 FPS	99.1%	122.0 W
2	Meridian Highway-Perception v3	NNVIDIA Jetson Thor	MIXED	4.09 ms	244 FPS	98.8%	91.6 W
3	YOLO26-Nano	QQualcomm QCS8550	INT8	4.71 ms	212 FPS	98.5%	12.9 W
4	NVIDIA-YOLO27-Orin	NNVIDIA Jetson AGX Orin	MIXED	6.02 ms	166 FPS	99.6%	48.8 W
5	RT-DETR-Edge	NNVIDIA Jetson Thor	FP16	6.12 ms	163 FPS	99.9%	102.3 W
6	YOLO26-Nano	AAmbarella CV72	INT8	6.25 ms	160 FPS	97.7%	5.2 W
7	YOLO27-Edge	NNVIDIA Jetson Thor	FP16	6.27 ms	160 FPS	99.6%	122.2 W
8	YOLO-World-Edge	NNVIDIA Jetson Thor	FP16	6.60 ms	151 FPS	99.9%	100.2 W
9	YOLO26-Nano	πRaspberry Pi 5 + Hailo HAT	INT8	7.70 ms	130 FPS	97.9%	6.9 W
10	Meridian Highway-Perception v3	NNVIDIA Jetson AGX Orin	MIXED	7.70 ms	130 FPS	99.0%	41.2 W
11	YOLO26-Nano	HHailo-8	INT8	8.28 ms	121 FPS	97.5%	2.3 W
12	RT-DETR-Edge	NNVIDIA Jetson AGX Orin	FP16	9.49 ms	105 FPS	99.8%	40.8 W
13	YOLO27-Edge	NNVIDIA Jetson AGX Orin	FP16	9.77 ms	102 FPS	99.4%	49.4 W
14	YOLO-World-Edge	NNVIDIA Jetson AGX Orin	FP16	11.4 ms	88 FPS	99.6%	52.3 W
15	YOLO26-Nano	QQualcomm QCS6490	INT8	12.5 ms	80 FPS	98.9%	8.2 W
16	YOLO27-Edge	QSnapdragon 8 Gen 3 NPU	INT8	13.1 ms	77 FPS	98.0%	6.3 W
17	YOLO27-Edge	QQualcomm QCS8550	INT8	13.9 ms	72 FPS	97.9%	13.0 W
18	YOLO26-Nano	AAmbarella CV5	INT8	14.3 ms	70 FPS	97.2%	4.0 W
19	EfficientDet-Lite4	πRaspberry Pi 5 + Hailo HAT	INT8	15.5 ms	64 FPS	98.4%	7.0 W
20	NVIDIA-YOLO27-Orin	NNVIDIA Jetson Orin Nano	INT8	15.5 ms	65 FPS	98.9%	13.1 W
21	Hailo-YOLO27-INT8	HHailo-10H	INT8	16.0 ms	62 FPS	97.9%	3.3 W
22	RT-DETR-Edge	QQualcomm QCS8550	INT8	16.1 ms	62 FPS	98.2%	13.1 W
23	Meridian Highway-Perception v3	QQualcomm QCS8550	INT8	16.5 ms	61 FPS	97.5%	12.9 W
24	Meridian Highway-Perception v3	HHailo-10H	INT8	16.8 ms	60 FPS	97.6%	3.3 W
25	YOLO27-Edge	HHailo-10H	INT8	17.0 ms	59 FPS	98.3%	3.3 W
26	YOLO-World-Edge	QQualcomm QCS8550	INT8	17.1 ms	58 FPS	98.5%	12.4 W
27	RT-DETR-Edge	NNVIDIA Jetson Orin Nano	INT8	17.3 ms	58 FPS	98.2%	11.4 W
28	YOLO27-Edge	NNVIDIA Jetson Orin Nano	INT8	18.1 ms	55 FPS	97.9%	12.2 W
29	YOLO27-Edge	AAMD Versal AI Edge VE2302	INT8	18.1 ms	55 FPS	97.4%	10.2 W
30	RT-DETR-Edge	AAMD Versal AI Edge VE2302	INT8	18.8 ms	53 FPS	98.9%	10.2 W
31	YOLO27-Edge	Apple Neural Engine (M4)	FP16	19.1 ms	52 FPS	99.9%	7.1 W
32	RT-DETR-Edge	Apple Neural Engine (M4)	FP16	19.6 ms	51 FPS	99.8%	6.8 W
33	YOLO26-Nano	RRockchip RK3588	INT8	19.8 ms	51 FPS	98.0%	9.3 W
34	YOLO27-Edge	AAmbarella CV72	INT8	20.7 ms	48 FPS	97.2%	5.5 W
35	YOLO27-Edge	HHailo-8	INT8	21.2 ms	47 FPS	97.5%	2.3 W
36	Hailo-YOLO27-INT8	πRaspberry Pi 5 + Hailo HAT	INT8	22.3 ms	45 FPS	98.9%	7.1 W
37	YOLO-World-Edge	NNVIDIA Jetson Orin Nano	INT8	22.6 ms	44 FPS	98.5%	13.1 W
38	Hailo-YOLO27-INT8	HHailo-8	INT8	22.7 ms	44 FPS	98.7%	2.2 W
39	YOLO26-Nano	GGoogle Coral Edge TPU	INT8	24.1 ms	42 FPS	97.8%	1.5 W
40	YOLO27-Edge	πRaspberry Pi 5 + Hailo HAT	INT8	25.9 ms	39 FPS	98.6%	7.4 W
41	YOLO26-Nano	IIntel Movidius Myriad X	INT8	26.3 ms	38 FPS	97.8%	2.4 W
42	EfficientDet-Lite4	QQualcomm QCS6490	INT8	27.0 ms	37 FPS	98.5%	8.3 W
43	EfficientDet-Lite4	AAmbarella CV5	INT8	27.1 ms	37 FPS	97.5%	3.9 W
44	YOLO-World-Edge	πRaspberry Pi 5 + Hailo HAT	INT8	35.0 ms	29 FPS	98.6%	7.3 W
45	EfficientDet-Lite4	RRockchip RK3588	INT8	43.8 ms	23 FPS	97.3%	8.9 W
46	EfficientDet-Lite4	GGoogle Coral Edge TPU	INT8	62.4 ms	16 FPS	98.8%	1.4 W
47	EfficientDet-Lite4	MMediaTek Genio 700	INT8	64.4 ms	16 FPS	98.3%	5.7 W
48	YOLO-World-Edge	RRockchip RK3588	INT8	69.0 ms	14 FPS	97.6%	8.2 W