Task object detection · model-size class nano (YOLO-n) · batch 1 · 16 platforms
✓ Like-for-like: same model-size class and batch.
| Platform | Result | Model · precision · stack | Grade | Power | Price | Fit perf/power/budget | Source |
|---|---|---|---|---|---|---|---|
| Hailo-8 (M.2/PCIe, x86 host) | 1036 FPS throughput b=1 | yolov8n INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) | D | 2.50 W reported in result source — spec | unknown | – / – / – | Hailo Model Zoo public models (Hailo-8) |
| NVIDIA DGX Spark | 990 FPS inference_time b=1 | YOLO11n FP16 · Ultralytics TensorRT | B | 153 W platform spec envelope (not measured with this result) | $4699 | – / – / – | Ultralytics docs benchmark table |
| Hailo-10H (M.2/PCIe, x86 host) | 375 FPS throughput b=1 | yolov8n INT8 (Hailo quantized) · Hailo DFC v5.4.0 / HailoRT (model zoo) | D | 2.50 W reported in result source — spec | unknown | – / – / – | Hailo Model Zoo public models (Hailo-10H) |
| NVIDIA Jetson AGX Thor | 719 FPS inference_time b=1 | YOLO26n FP16 · Ultralytics TensorRT | B | 130 W platform spec envelope (not measured with this result) | $3499 | – / – / – | Ultralytics docs benchmark table |
| Apple M4 (10-core GPU, 120 GB/s) | 357 FPS latency_e2e b=1 | YOLO26n FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 | B | unknown no power published | unknown | – / – / – | Ultralytics docs: Apple Core AI integration benchmark |
| Intel Core Ultra 7 258V (Lunar Lake) | 300 FPS inference_time b=1 | YOLO26n FP32 · Ultralytics + OpenVINO 2026.2.1 | C | unknown no power published | unknown | – / – / – | Ultralytics docs: Intel OpenVINO integration benchmark |
| NVIDIA Jetson Orin NX 16GB | 287 FPS inference_time b=1 | YOLO26n INT8 · Ultralytics TensorRT | B | unknown no power published | unknown | – / – / – | Ultralytics docs benchmark table |
| Hailo-8L (M.2/PCIe, x86 host) | 201 FPS throughput b=1 | yolov8n INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) | D | 1.50 W reported in result source — spec | unknown | – / – / – | Hailo Model Zoo public models (Hailo-8L) |
| Intel Core Ultra X7 358H (Panther Lake) | 248 FPS inference_time b=1 | YOLO26n FP16 · Ultralytics + OpenVINO 2026.2.1 | C | unknown no power published | unknown | – / – / – | Ultralytics docs: Intel OpenVINO integration benchmark |
| AMD Ryzen AI Max+ 395 / PRO 395 (Strix Halo, Radeon 8060S) | 476 FPS inference_time b=1 | YOLO26n FP32 · Ultralytics 8.4.165 ONNX Runtime MIGraphX EP | B | 120 W platform spec envelope (not measured with this result) | unknown | – / – / – | Ultralytics docs: AMD ROCm/MIGraphX integration benchmark |
| NVIDIA Jetson AGX Orin 64GB | 435 FPS inference_time b=1 | YOLO26n INT8 · Ultralytics TensorRT | B | 65.00 W platform spec envelope (not measured with this result) | $1999 | – / – / – | Ultralytics docs benchmark table |
| Intel Core Ultra 7 155H (Meteor Lake) | 173 FPS inference_time b=1 | YOLO26n INT8 · Ultralytics + OpenVINO 2026.2.1 | C | unknown no power published | unknown | – / – / – | Ultralytics docs: Intel OpenVINO integration benchmark |
| NVIDIA Jetson Orin Nano Super | 263 FPS inference_time b=1 | YOLO26n INT8 · Ultralytics TensorRT | B | 25.00 W reported in result source — spec | $249 | – / – / – | Ultralytics docs benchmark table |
| Microchip PolarFire SoC Video Kit (VectorBlox CoreVectorBlox) | 66.21 FPS inference_time b=1 | yolov5n_512x288 INT8 (VNNX) · VectorBlox SDK (VNNX), no compression | D | unknown no power published | unknown | – / – / – | Microchip VectorBlox SDK tutorial metrics |
| Raspberry Pi 5 + AI HAT+ (Hailo-8) | 46.08 FPS inference_time b=1 | YOLO26n INT8 · Ultralytics Hailo HEF / HailoRT | B | unknown no power published | unknown | – / – / – | Ultralytics docs: Hailo integration benchmark |
| Raspberry Pi 5 (CPU only) | 14.92 FPS inference_time b=1 | YOLO26n FP32/default · Ultralytics NCNN | B | unknown no power published | unknown | – / – / – | Ultralytics docs benchmark table |
Confidence: A audited MLPerf closed, B documented third-party, C community, D vendor-published, E estimate. Unknown is never assumed. Other like-for-like groups: nano (YOLO-n) b1 (16); small (YOLO-s) b1 (8); medium (YOLO-m) b1 (8); medium (YOLO-m) b8 (4); large (YOLO-l) b1 (4); xlarge (YOLO-x) b1 (4); nano (YOLO-n) b8 (3); small (YOLO-s) b8 (3); mobile-class CNN b1 (1); mid-size CNN bunknown (34); medium (YOLO-m) bunknown (6); mobile-class CNN bunknown (6); small (YOLO-s) bunknown (4); unknown b1 (7); unknown bunknown (37)