image classification benchmarks on edge AI hardware
Each table compares only results with the same model-size class and batch size. Search this workload · JSON.
Fastest in the best-covered like-for-like group (mobile-class CNN, batch 1): Qualcomm Snapdragon X2 Elite (CRD) at 3953 FPS D 2026 (year only). 13 groups, 61 platforms.
Show data — 13 like-for-like tables
mobile-class CNN · batch 1 (16 platforms)
✓ Like-for-like: model-size class mobile-class CNN, batch 1. Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Qualcomm Snapdragon X2 Elite (CRD) | 3953 FPS inference_time · b=1 | mobilenet_v2 w8a16 · Qualcomm AI Hub onnx (QAIRT 2.50.0.260828221209) · mobile-class CNN | D | unknown power n/a no power published | Qualcomm AI Hub Models perf.yaml (BSD-3) 2026 (year only) |
| Hailo-10H (M.2/PCIe, x86 host) | 2798 FPS throughput · b=1 | mobilenet_v2_1.0 INT8 (Hailo quantized) · Hailo DFC v5.4.0 / HailoRT (model zoo) · mobile-class CNN | D | 2.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-10H) undated |
| Hailo-8 (M.2/PCIe, x86 host) | 2597 FPS throughput · b=1 | mobilenet_v2_1.0 INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) · mobile-class CNN | D | 2.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-8) undated |
| Qualcomm Snapdragon X Elite (CRD) | 2203 FPS inference_time · b=1 | mobilenet_v2 w8a8 · Qualcomm AI Hub onnx (QAIRT 2.50.0.260828221209) · mobile-class CNN | D | unknown power n/a no power published | Qualcomm AI Hub Models perf.yaml (BSD-3) 2026 (year only) |
| Qualcomm Dragonwing IQ-9075 EVK | 1965 FPS inference_time · b=1 | mobilenet_v2 w8a8 · Qualcomm AI Hub qnn_dlc (QAIRT 2.50.0.260828221209) · mobile-class CNN | D | unknown power n/a no power published | Qualcomm AI Hub Models perf.yaml (BSD-3) 2026 (year only) |
| Hailo-8L (M.2/PCIe, x86 host) | 1738 FPS throughput · b=1 | mobilenet_v2_1.0 INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) · mobile-class CNN | D | 1.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-8L) undated |
| Qualcomm Dragonwing RB3 Gen 2 (QCS6490) | 940 FPS inference_time · b=1 | mobilenet_v2 w8a8 · Qualcomm AI Hub onnx (QAIRT 2.50.0.260828221209) · mobile-class CNN | D | unknown power n/a no power published | Qualcomm AI Hub Models perf.yaml (BSD-3) 2026 (year only) |
| Microchip PolarFire SoC Video Kit (VectorBlox CoreVectorBlox) | 672 FPS inference_time · b=1 | mobilenet-v1-0.25-128 INT8 (VNNX) · VectorBlox SDK (VNNX), no compression · mobile-class CNN | D | unknown power n/a no power published | Microchip VectorBlox SDK tutorial metrics undated |
| NXP i.MX 9 with eIQ Neutron-S NPU (i.MX 95/952/943; doc does not specify) | 619 FPS inference_time · b=1 | mobilenetv2-12 (neutron-converted ONNX) INT8 (quantized) · ONNX Runtime 1.24.3 onnxruntime_perf_test; delegate: Neutron execution provider · mobile-class CNN | D | unknown power n/a no power published | NXP UG10166 (vendor document) undated |
| NXP i.MX 95 | 588 FPS inference_time · b=1 | mobilenet_v1_1.0_224_quant INT8 (quantized) · TensorFlow Lite 2.19.0 label_image.py (Python); delegate: Neutron Delegate · mobile-class CNN | D | unknown power n/a no power published | NXP UG10166 (vendor document) undated |
| Google Coral USB Accelerator (Edge TPU) on Xeon Gold 6154 host, USB 3.0 | 417 FPS inference_time · b=1 | MobileNet v1 (224x224) INT8 · TensorFlow Lite C++ benchmark (Edge TPU runtime) · mobile-class CNN | D | 2.00 W spec platform spec envelope (not measured with this result) | Google Coral Edge TPU benchmarks page (Table 1) undated |
| Google Coral Dev Board (Edge TPU) | 417 FPS inference_time · b=1 | MobileNet v1 (224x224) INT8 · TensorFlow Lite C++ benchmark (Edge TPU runtime) · mobile-class CNN | D | 2.00 W spec platform spec envelope (not measured with this result) | Google Coral Edge TPU benchmarks page (Table 1) undated |
| NXP i.MX 8M Plus | 390 FPS inference_time · b=1 | mobilenet_v1_1.0_224_quant INT8 (quantized) · TensorFlow Lite 2.19.0 label_image (C++); delegate: VX Delegate (USE_GPU_INFERENCE=0) · mobile-class CNN | D | unknown power n/a no power published | NXP UG10166 (vendor document) undated |
| NXP i.MX 93 | 256 FPS inference_time · b=1 | mobilenet_v1_1.0_224_quant (vela) INT8 (quantized) · TensorFlow Lite 2.19.0 label_image (C++); delegate: Ethos-U Delegate (Vela-compiled model) · mobile-class CNN | D | unknown power n/a no power published | NXP UG10166 (vendor document) undated |
| Intel Xeon Gold 6154 desktop (CPU only, TFLite) CPU-only | 19.61 FPS inference_time · b=1 | MobileNet v2 (224x224) FP32/INT8 TFLite on CPU · TensorFlow Lite C++ benchmark (Edge TPU runtime) · mobile-class CNN | D | unknown power n/a no power published | Google Coral Edge TPU benchmarks page (Table 1) undated |
| Coral Dev Board CPU (quad Cortex-A53 @1.5 GHz, CPU only) CPU-only | 8.20 FPS inference_time · b=1 | MobileNet v2 (224x224) TFLite on CPU · TensorFlow Lite C++ benchmark (Edge TPU runtime) · mobile-class CNN | D | unknown power n/a no power published | Google Coral Edge TPU benchmarks page (Table 1) undated |
mid-size CNN · batch 1 (13 platforms)
✓ Like-for-like: model-size class mid-size CNN, batch 1. Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Qualcomm Snapdragon X2 Elite (CRD) | 2445 FPS inference_time · b=1 | resnet50 w8a8 · Qualcomm AI Hub onnx (QAIRT 2.50.0.260828221209) · mid-size CNN | D | unknown power n/a no power published | Qualcomm AI Hub Models perf.yaml (BSD-3) 2026 (year only) |
| Hailo-8 (M.2/PCIe, x86 host) | 1372 FPS throughput · b=1 | resnet_v1_50 INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) · mid-size CNN | D | 2.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-8) undated |
| Qualcomm Dragonwing IQ-9075 EVK | 1192 FPS inference_time · b=1 | resnet50 w8a8 · Qualcomm AI Hub tflite (QAIRT 2.50.0.260828221209) · mid-size CNN | D | unknown power n/a no power published | Qualcomm AI Hub Models perf.yaml (BSD-3) 2026 (year only) |
| Qualcomm Snapdragon X Elite (CRD) | 1186 FPS inference_time · b=1 | resnet50 w8a8 · Qualcomm AI Hub onnx (QAIRT 2.50.0.260828221209) · mid-size CNN | D | unknown power n/a no power published | Qualcomm AI Hub Models perf.yaml (BSD-3) 2026 (year only) |
| Qualcomm Dragonwing RB3 Gen 2 (QCS6490) | 379 FPS inference_time · b=1 | resnet50 w8a8 · Qualcomm AI Hub onnx (QAIRT 2.50.0.260828221209) · mid-size CNN | D | unknown power n/a no power published | Qualcomm AI Hub Models perf.yaml (BSD-3) 2026 (year only) |
| Hailo-10H (M.2/PCIe, x86 host) | 331 FPS throughput · b=1 | resnet_v1_50 INT8 (Hailo quantized) · Hailo DFC v5.4.0 / HailoRT (model zoo) · mid-size CNN | D | 2.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-10H) undated |
| Google Coral USB Accelerator (Edge TPU) on Xeon Gold 6154 host, USB 3.0 | 303 FPS inference_time · b=1 | Unet Mv2 (128x128) INT8 · TensorFlow Lite C++ benchmark (Edge TPU runtime) · mid-size CNN | D | 2.00 W spec platform spec envelope (not measured with this result) | Google Coral Edge TPU benchmarks page (Table 1) undated |
| Altera Agilex 7 FPGA I-Series Development Kit (FPGA AI Suite) | 182 FPS throughput · b=1 | ResNet18 (PyTorch) not stated in table · FPGA AI Suite on-chip parameter streaming; IP throughput · mid-size CNN | D | unknown power n/a no power published | Altera FPGA AI Suite Handbook (performance tables) undated |
| Google Coral Dev Board (Edge TPU) | 175 FPS inference_time · b=1 | Unet Mv2 (128x128) INT8 · TensorFlow Lite C++ benchmark (Edge TPU runtime) · mid-size CNN | D | 2.00 W spec platform spec envelope (not measured with this result) | Google Coral Edge TPU benchmarks page (Table 1) undated |
| Hailo-8L (M.2/PCIe, x86 host) | 138 FPS throughput · b=1 | resnet_v1_50 INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) · mid-size CNN | D | 1.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-8L) undated |
| Microchip PolarFire SoC Video Kit (VectorBlox CoreVectorBlox) | 68.35 FPS inference_time · b=1 | resnet18_86s_07p INT8 (VNNX) · VectorBlox SDK (VNNX), unstructured compression · mid-size CNN | D | unknown power n/a no power published | Microchip VectorBlox SDK tutorial metrics undated |
| Intel Xeon Gold 6154 desktop (CPU only, TFLite) CPU-only | 36.10 FPS inference_time · b=1 | Unet Mv2 (128x128) FP32/INT8 TFLite on CPU · TensorFlow Lite C++ benchmark (Edge TPU runtime) · mid-size CNN | D | unknown power n/a no power published | Google Coral Edge TPU benchmarks page (Table 1) undated |
| Coral Dev Board CPU (quad Cortex-A53 @1.5 GHz, CPU only) CPU-only | 5.24 FPS inference_time · b=1 | Unet Mv2 (128x128) TFLite on CPU · TensorFlow Lite C++ benchmark (Edge TPU runtime) · mid-size CNN | D | unknown power n/a no power published | Google Coral Edge TPU benchmarks page (Table 1) undated |
nano (YOLO-n) · batch 1 (4 platforms)
✓ Like-for-like: model-size class nano (YOLO-n), batch 1. Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Raspberry Pi 5 + AI HAT+ (Hailo-8) | 500 FPS inference_time · b=1 | YOLO26n-cls INT8 · Ultralytics Hailo HEF / HailoRT · nano (YOLO-n) | B | unknown power n/a no power published | Ultralytics docs: Hailo integration benchmark 2026 (year only) |
| Apple M4 (10-core GPU, 120 GB/s) | 417 FPS latency_e2e · b=1 | YOLO26n-cls FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n) | B | unknown power n/a no power published | Ultralytics docs: Apple Core AI integration benchmark 2026 (year only) |
| Microchip PolarFire SoC Video Kit (VectorBlox CoreVectorBlox) | 244 FPS inference_time · b=1 | yolov8n-cls INT8 (VNNX) · VectorBlox SDK (VNNX), no compression · nano (YOLO-n) | D | unknown power n/a no power published | Microchip VectorBlox SDK tutorial metrics undated |
| Raspberry Pi 5 (CPU only) CPU-only | 29.59 FPS inference_time · b=1 | YOLO26n-cls FP32 · Ultralytics PyTorch (CPU) · nano (YOLO-n) | B | unknown power n/a no power published | Ultralytics docs: Hailo integration benchmark 2026 (year only) |
mobile-class CNN · batch 8 (3 platforms)
✓ Like-for-like: model-size class mobile-class CNN, batch 8. Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Hailo-10H (M.2/PCIe, x86 host) | 2753 FPS throughput · b=8 | mobilenet_v2_1.0 INT8 (Hailo quantized) · Hailo DFC v5.4.0 / HailoRT (model zoo) · mobile-class CNN | D | 2.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-10H) undated |
| Hailo-8 (M.2/PCIe, x86 host) | 2597 FPS throughput · b=8 | mobilenet_v2_1.0 INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) · mobile-class CNN | D | 2.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-8) undated |
| Hailo-8L (M.2/PCIe, x86 host) | 1738 FPS throughput · b=8 | mobilenet_v2_1.0 INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) · mobile-class CNN | D | 1.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-8L) undated |
mid-size CNN · batch 8 (3 platforms)
✓ Like-for-like: model-size class mid-size CNN, batch 8. Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Hailo-8 (M.2/PCIe, x86 host) | 1372 FPS throughput · b=8 | resnet_v1_50 INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) · mid-size CNN | D | 2.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-8) undated |
| Hailo-10H (M.2/PCIe, x86 host) | 1024 FPS throughput · b=8 | resnet_v1_50 INT8 (Hailo quantized) · Hailo DFC v5.4.0 / HailoRT (model zoo) · mid-size CNN | D | 2.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-10H) undated |
| Hailo-8L (M.2/PCIe, x86 host) | 502 FPS throughput · b=8 | resnet_v1_50 INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) · mid-size CNN | D | 1.50 W spec reported in result source — spec | Hailo Model Zoo public models (Hailo-8L) undated |
large (YOLO-l) · batch 1 (1 platforms)
✓ Like-for-like: model-size class large (YOLO-l), batch 1. Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Microchip PolarFire SoC Video Kit (VectorBlox CoreVectorBlox) | 10.43 FPS inference_time · b=1 | yolov8l-cls INT8 (VNNX) · VectorBlox SDK (VNNX), no compression · large (YOLO-l) | D | unknown power n/a no power published | Microchip VectorBlox SDK tutorial metrics undated |
medium (YOLO-m) · batch 1 (1 platforms)
✓ Like-for-like: model-size class medium (YOLO-m), batch 1. Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Microchip PolarFire SoC Video Kit (VectorBlox CoreVectorBlox) | 27.05 FPS inference_time · b=1 | yolov8m-cls INT8 (VNNX) · VectorBlox SDK (VNNX), no compression · medium (YOLO-m) | D | unknown power n/a no power published | Microchip VectorBlox SDK tutorial metrics undated |
small (YOLO-s) · batch 1 (1 platforms)
✓ Like-for-like: model-size class small (YOLO-s), batch 1. Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Microchip PolarFire SoC Video Kit (VectorBlox CoreVectorBlox) | 86.81 FPS inference_time · b=1 | yolov8s-cls INT8 (VNNX) · VectorBlox SDK (VNNX), no compression · small (YOLO-s) | D | unknown power n/a no power published | Microchip VectorBlox SDK tutorial metrics undated |
xlarge (YOLO-x) · batch 1 (1 platforms)
✓ Like-for-like: model-size class xlarge (YOLO-x), batch 1. Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Microchip PolarFire SoC Video Kit (VectorBlox CoreVectorBlox) | 3.67 FPS inference_time · b=1 | yolov8x-cls INT8 (VNNX) · VectorBlox SDK (VNNX), no compression · xlarge (YOLO-x) | D | unknown power n/a no power published | Microchip VectorBlox SDK tutorial metrics undated |
mid-size CNN · batch unknown (41 platforms)
✓ Like-for-like: model-size class mid-size CNN, batch unknown (size or batch not stated by source — unverified). Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| HPE ProLiant Compute DL380 Gen12 (5x NVIDIA RTX PRO 4500 Blackwell) | 94651 FPS throughput_offline · b=unknown | MLPerf resnet int8 · TensorRT 10.11.0.33, CUDA 13.0 · mid-size CNN | A | unknown power n/a no power published | MLPerf Inference v6.0 (closed, edge) ID 6.0-0056 undated |
| GATE Overflow workstation, Intel Xeon w7-2495X (2x NVIDIA GeForce RTX 4090) | 87946 FPS throughput_offline · b=unknown | MLPerf resnet int8 · TensorRT · mid-size CNN | A | unknown power n/a no power published | MLPerf Inference v6.0 (closed, edge) ID 6.0-0035 undated |
| Intel Xeon Platinum 6972P CPU-only | 28295 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Lenovo ThinkEdge SE455i V3 (2x NVIDIA L4-PCIe-24GB) | 25201 FPS throughput_offline · b=unknown | MLPerf resnet int8 · TensorRT 10.8, CUDA 13.1 · mid-size CNN | A | unknown power n/a no power published | MLPerf Inference v6.0 (closed, edge) ID 6.0-0068 undated |
| Intel Xeon Platinum 8580 CPU-only | 21025 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| HPE ProLiant DL145 Gen11 (1x NVIDIA RTX PRO 4500 Blackwell) | 19211 FPS throughput_offline · b=unknown | MLPerf resnet int8 · TensorRT 10.11.0.33, CUDA 13.0 · mid-size CNN | A | unknown power n/a no power published | MLPerf Inference v6.0 (closed, edge) ID 6.0-0055 undated |
| Intel Xeon Platinum 8480+ CPU-only | 18307 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| AMD Alveo V70 (Versal AI Core, XDNA AIE-ML) | 7780 FPS E2E throughput, multi thread · b=unknown | pt_OFA-resnet50_0.88_3.5 INT8 · Vitis AI 3.5 (VART) · mid-size CNN | D | 75.00 W spec reported in result source — spec | Vitis AI 3.5 Model Zoo performance table Older data (2023) |
| AMD Versal AI Edge VEK280 (VE2802) | 6595 FPS E2E throughput, multi thread · b=unknown | pt_OFA-resnet50_0.88_3.5 INT8 · Vitis AI 3.5 (VART) · mid-size CNN | D | unknown power n/a no power published | Vitis AI 3.5 Model Zoo performance table Older data (2023) |
| NVIDIA Jetson AGX Orin 64GB | 6424 FPS throughput_offline · b=unknown | MLPerf ResNet INT8/FP16 (MLPerf closed) · TensorRT (MLPerf Inference, older round) · mid-size CNN | D | 65.00 W spec platform spec envelope (not measured with this result) | NVIDIA Jetson Benchmarks (MLPerf v3.1 results table) 2026-10 |
| Intel Xeon Platinum 8380 CPU-only | 4762 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| AMD Versal AI Core VCK190 (VC1902) | 4566 FPS E2E throughput, multi thread · b=unknown | tf_resnetv1_50_imagenet_224_224_0.65_2.45G INT8 · Vitis AI 3.0 (VART) · mid-size CNN | D | unknown power n/a no power published | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| NVIDIA Jetson AGX Orin (MaxQ) | 3526 FPS throughput_offline · b=unknown | MLPerf ResNet INT8/FP16 (MLPerf closed) · TensorRT (MLPerf Inference, older round) · mid-size CNN | D | 23.57 W measured reported in result source — measured | NVIDIA Jetson Benchmarks (MLPerf v3.1 results table) 2026-10 |
| Intel Xeon Gold 6338N CPU-only | 3321 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Xeon Platinum 8280 CPU-only | 2937 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| NVIDIA Jetson Orin NX 16GB | 2641 FPS throughput_offline · b=unknown | MLPerf ResNet INT8/FP16 (MLPerf closed) · TensorRT (MLPerf Inference, older round) · mid-size CNN | D | unknown power n/a no power published | NVIDIA Jetson Benchmarks (MLPerf v3.1 results table) 2026-10 |
| Intel Data Center GPU Flex 170 | 2106 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core Ultra 9 288V | 1981 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Arc B-Series Graphics | 1926 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 FP16 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| NVIDIA Jetson Orin NX 16GB (MaxQ) | 1682 FPS throughput_offline · b=unknown | MLPerf ResNet INT8/FP16 (MLPerf closed) · TensorRT (MLPerf Inference, older round) · mid-size CNN | D | 14.95 W measured reported in result source — measured | NVIDIA Jetson Benchmarks (MLPerf v3.1 results table) 2026-10 |
| Intel Arc A-Series Graphics | 1357 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core Ultra 7 265H CPU-only | 1121 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core Ultra 7 155H CPU-only | 980 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Xeon Gold 5218T CPU-only | 958 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core Ultra 7 255H (Lenovo ThinkEdge SE100) | 835 FPS throughput_offline · b=unknown | MLPerf resnet INT8 · OpenVINO · mid-size CNN | A | unknown power n/a no power published | MLPerf Inference v6.0 (closed, edge) ID 6.0-0067 undated |
| HPE ProLiant Compute DL380 Gen12 (1x NVIDIA RTX PRO 4500 Blackwell-PCIe-32GB) | 675 FPS latency_single_stream · b=unknown | MLPerf resnet50 fp32 · ONNX Runtime v1.28.0 · mid-size CNN | B | unknown power n/a no power published | MLPerf Inference v6.1 (open, edge) ID HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT_open undated |
| Intel Core i5-13600K CPU-only | 531 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i7-12700H CPU-only | 522 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| AMD Zynq UltraScale+ ZCU102 (ZU9EG) | 371 FPS E2E throughput, multi thread · b=unknown | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G INT8 · Vitis AI 3.0 (VART) · mid-size CNN | D | unknown power n/a no power published | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| Intel Core i7-1185G7 CPU-only | 358 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| AMD Zynq UltraScale+ ZCU104 (ZU7EV) | 354 FPS E2E throughput, multi thread · b=unknown | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G INT8 · Vitis AI 3.0 (VART) · mid-size CNN | D | unknown power n/a no power published | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| Intel Core i7-1355U CPU-only | 335 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i5-1335U CPU-only | 231 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i7-1185GRE CPU-only | 223 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| AMD Kria KV260 Vision AI Starter Kit (K26 SOM) | 214 FPS E2E throughput, multi thread · b=unknown | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G INT8 · Vitis AI 3.0 (VART) · mid-size CNN | D | 7.50 W spec reported in result source — spec | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| Intel Core i5-1235U | 200 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Celeron 6305E | 199 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Atom x7425E CPU-only | 126 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel N100 CPU-only | 110 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| TRIGKEY Mini PC (AMD Ryzen 5 5560U, CPU) CPU-only | 104 FPS throughput_offline · b=unknown | MLPerf resnet (open) / efficientnet-int8-lite0 INT8 · MLPerf tflite vmaster (ctuning_cpp_tflite-cpu-tflite_vmaster-default_config) · mid-size CNN | B | 28.20 W measured reported in result source — measured | MLPerf Inference v5.0 (open) ID 5.0-5436 2025 (year only) |
| Intel Atom X6425E CPU-only | 58.96 FPS throughput (OpenVINO benchmark_app) · b=unknown | resnet-50 INT8 · OpenVINO 2026.4.1 · mid-size CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
mobile-class CNN · batch unknown (31 platforms)
✓ Like-for-like: model-size class mobile-class CNN, batch unknown (size or batch not stated by source — unverified). Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| AMD Alveo V70 (Versal AI Core, XDNA AIE-ML) | 63109 FPS E2E throughput, multi thread · b=unknown | tf_mobilenetv1-0.25_3.5 INT8 · Vitis AI 3.5 (VART) · mobile-class CNN | D | 75.00 W spec reported in result source — spec | Vitis AI 3.5 Model Zoo performance table Older data (2023) |
| Intel Xeon Platinum 8580 CPU-only | 39672 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Xeon Platinum 8480+ CPU-only | 37258 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Xeon Platinum 6972P CPU-only | 36319 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Xeon Platinum 8380 CPU-only | 22335 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Xeon Gold 6338N CPU-only | 16247 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Xeon Platinum 8280 CPU-only | 14815 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| AMD Versal AI Core VCK190 (VC1902) | 10225 FPS E2E throughput, multi thread · b=unknown | tf_mobilenetv1_0.25_imagenet_128_128_27M INT8 · Vitis AI 3.0 (VART) · mobile-class CNN | D | unknown power n/a no power published | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| AMD Versal AI Edge VEK280 (VE2802) | 8381 FPS E2E throughput, multi thread · b=unknown | tf_mobilenetv1-0.25_3.5 INT8 · Vitis AI 3.5 (VART) · mobile-class CNN | D | unknown power n/a no power published | Vitis AI 3.5 Model Zoo performance table Older data (2023) |
| Intel Xeon Gold 5218T CPU-only | 5447 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core Ultra 7 155H CPU-only | 4952 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| AMD Zynq UltraScale+ ZCU102 (ZU9EG) | 4832 FPS E2E throughput, multi thread · b=unknown | tf_mobilenetv1_0.25_imagenet_128_128_27M INT8 · Vitis AI 3.0 (VART) · mobile-class CNN | D | unknown power n/a no power published | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| AMD Zynq UltraScale+ ZCU104 (ZU7EV) | 4325 FPS E2E throughput, multi thread · b=unknown | tf_mobilenetv1_0.25_imagenet_128_128_27M INT8 · Vitis AI 3.0 (VART) · mobile-class CNN | D | unknown power n/a no power published | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| Intel Core Ultra 9 288V | 4065 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Data Center GPU Flex 170 | 3199 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 FP16 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| TRIGKEY Mini PC (AMD Ryzen 5 5560U, CPU) CPU-only | 2798 FPS throughput_offline · b=unknown | MLPerf resnet (open) / mobilenet-v1-precision_float-0.25-128 FP32 · MLPerf tflite vmaster (ctuning_cpp_tflite-cpu-tflite_vmaster-default_config) · mobile-class CNN | B | 30.72 W measured reported in result source — measured | MLPerf Inference v5.0 (open) ID 5.0-5441 2025 (year only) |
| AMD Kria KV260 Vision AI Starter Kit (K26 SOM) | 2432 FPS E2E throughput, multi thread · b=unknown | tf_mobilenetv1_0.25_imagenet_128_128_27M INT8 · Vitis AI 3.0 (VART) · mobile-class CNN | D | 7.50 W spec reported in result source — spec | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| Intel Arc A-Series Graphics | 2309 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core Ultra 7 265H CPU-only | 2271 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Arc B-Series Graphics | 2154 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 FP16 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i7-12700H CPU-only | 2099 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i7-1185G7 CPU-only | 1344 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i5-13600K CPU-only | 1275 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 FP32 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i7-1355U CPU-only | 1267 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i5-1335U CPU-only | 1033 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i7-1185GRE CPU-only | 966 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Core i5-1235U CPU-only | 755 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Celeron 6305E CPU-only | 506 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel N100 CPU-only | 478 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Atom x7425E CPU-only | 444 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
| Intel Atom X6425E CPU-only | 226 FPS throughput (OpenVINO benchmark_app) · b=unknown | mobilenet-v2 INT8 · OpenVINO 2026.4.1 · mobile-class CNN | D | unknown power n/a no power published | OpenVINO official performance benchmarks (graph-data-ov.json) undated |
unknown · batch 1 (5 platforms)
✓ Like-for-like: model-size class unknown, batch 1 (size or batch not stated by source — unverified). Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Google Coral Dev Board (Edge TPU) | 500 FPS inference_time · b=1 | SqueezeNet (224x224) INT8 · TensorFlow Lite C++ benchmark (Edge TPU runtime) · unknown | D | 2.00 W spec platform spec envelope (not measured with this result) | Google Coral Edge TPU benchmarks page (Table 1) undated |
| Google Coral USB Accelerator (Edge TPU) on Xeon Gold 6154 host, USB 3.0 | 476 FPS inference_time · b=1 | SqueezeNet (224x224) INT8 · TensorFlow Lite C++ benchmark (Edge TPU runtime) · unknown | D | 2.00 W spec platform spec envelope (not measured with this result) | Google Coral Edge TPU benchmarks page (Table 1) undated |
| Microchip PolarFire SoC Video Kit (VectorBlox CoreVectorBlox) | 99.88 FPS inference_time · b=1 | onnx_squeezenet1.1 INT8 (VNNX) · VectorBlox SDK (VNNX), no compression · unknown | D | unknown power n/a no power published | Microchip VectorBlox SDK tutorial metrics undated |
| Intel Xeon Gold 6154 desktop (CPU only, TFLite) CPU-only | 18.18 FPS inference_time · b=1 | SqueezeNet (224x224) FP32/INT8 TFLite on CPU · TensorFlow Lite C++ benchmark (Edge TPU runtime) · unknown | D | unknown power n/a no power published | Google Coral Edge TPU benchmarks page (Table 1) undated |
| Coral Dev Board CPU (quad Cortex-A53 @1.5 GHz, CPU only) CPU-only | 4.31 FPS inference_time · b=1 | SqueezeNet (224x224) TFLite on CPU · TensorFlow Lite C++ benchmark (Edge TPU runtime) · unknown | D | unknown power n/a no power published | Google Coral Edge TPU benchmarks page (Table 1) undated |
unknown · batch unknown (6 platforms)
✓ Like-for-like: model-size class unknown, batch unknown (size or batch not stated by source — unverified). Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| AMD Alveo V70 (Versal AI Core, XDNA AIE-ML) | 12926 FPS E2E throughput, multi thread · b=unknown | pt_OFA-depthwise-res50_3.5 INT8 · Vitis AI 3.5 (VART) · unknown | D | 75.00 W spec reported in result source — spec | Vitis AI 3.5 Model Zoo performance table Older data (2023) |
| AMD Versal AI Edge VEK280 (VE2802) | 8598 FPS E2E throughput, multi thread · b=unknown | pt_squeezenet_3.5 INT8 · Vitis AI 3.5 (VART) · unknown | D | unknown power n/a no power published | Vitis AI 3.5 Model Zoo performance table Older data (2023) |
| AMD Versal AI Core VCK190 (VC1902) | 5827 FPS E2E throughput, multi thread · b=unknown | pt_squeezenet_imagenet_224_224_703.5M INT8 · Vitis AI 3.0 (VART) · unknown | D | unknown power n/a no power published | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| AMD Zynq UltraScale+ ZCU102 (ZU9EG) | 1551 FPS E2E throughput, multi thread · b=unknown | pt_squeezenet_imagenet_224_224_703.5M INT8 · Vitis AI 3.0 (VART) · unknown | D | unknown power n/a no power published | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| AMD Zynq UltraScale+ ZCU104 (ZU7EV) | 1326 FPS E2E throughput, multi thread · b=unknown | pt_squeezenet_imagenet_224_224_703.5M INT8 · Vitis AI 3.0 (VART) · unknown | D | unknown power n/a no power published | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
| AMD Kria KV260 Vision AI Starter Kit (K26 SOM) | 760 FPS E2E throughput, multi thread · b=unknown | pt_squeezenet_imagenet_224_224_703.5M INT8 · Vitis AI 3.0 (VART) · unknown | D | 7.50 W spec reported in result source — spec | Vitis AI 3.0 Model Zoo performance table Older data (2023) |
How to read these numbers
Every number is labelled. Confidence: A MLPerf closed, audited · B documented third-party method · C community / review · D vendor-published, not independently verified · E estimate. Power basis: measured with the result · spec vendor spec/envelope · power n/a not published.
Machine-readable: /find.json (see llms.txt); full tables are in “Show data”. Confidence grades: A audited MLPerf closed · B documented third-party · C community · D vendor-published · E estimate. Unknown values are shown as unknown, never estimated. Generated 2026-10-16 from public sources.