AMD Kria KV260 Vision AI Starter Kit (K26 SOM): edge AI benchmarks

Name: AMD Kria KV260 Vision AI Starter Kit (K26 SOM)
Brand: AMD

FPGA / Versal coverage note: 1,243 rows come from the AMD Vitis AI 3.0/3.5 Model Zoo tables; 733 (59%) are image classification. Detection, segmentation and other workloads are thinner, and there is no LLM or speech data for these boards. All are vendor-published (grade D). FPGA / Versal overview.

AMDSOM starter kit (adaptive SoC)No licensed photo yet

Vendor AMDForm factor SOM starter kit (adaptive SoC)Compute DPU (programmable logic)Indicative price unknownPower envelope 7.5 W
spec: K26 SOM typical 7.5 W (max 15 W), carrier board not included (Kria PSG)Published rows 274

Key results

Best published result per workload for AMD Kria KV260 Vision AI Starter Kit (K26 SOM)
Workload	Best result	Model · precision	Power · perf/W	Grade · date	Source
object detection	276 FPS batch unknown	pt_yolox-nano_coco_416_416_1G INT8 · Vitis AI 3.0 (VART)	7.50 W (spec) 36.76 FPS/W	D Older data (2023)	Vitis AI 3.0 Model Zoo performance table
pose estimation	352 FPS batch unknown	pt_movenet_coco_192_192_0.5G INT8 · Vitis AI 3.0 (VART)	7.50 W (spec) 46.97 FPS/W	D Older data (2023)	Vitis AI 3.0 Model Zoo performance table
image classification	2432 FPS batch unknown	tf_mobilenetv1_0.25_imagenet_128_128_27M INT8 · Vitis AI 3.0 (VART)	7.50 W (spec) 324 FPS/W	D Older data (2023)	Vitis AI 3.0 Model Zoo performance table
semantic segmentation	183 FPS batch unknown	tf2_2d-unet_nuclei_128_128_5.31G INT8 · Vitis AI 3.0 (VART)	7.50 W (spec) 24.39 FPS/W	D Older data (2023)	Vitis AI 3.0 Model Zoo performance table
depth estimation	4.51 FPS batch unknown	pt_fadnetv2_sceneflow_576_960_0.51_201G INT8 · Vitis AI 3.0 (VART)	7.50 W (spec) 0.601 FPS/W	D Older data (2023)	Vitis AI 3.0 Model Zoo performance table
super-resolution	105 FPS batch unknown	pt_SESR-S_DIV2K_360_640_10.2G INT8 · Vitis AI 3.0 (VART)	7.50 W (spec) 14.05 FPS/W	D Older data (2023)	Vitis AI 3.0 Model Zoo performance table
keypoint / feature detection	11.31 FPS batch unknown	tf_HFNet_mixed_960_960_20.09G INT8 · Vitis AI 3.0 (VART)	7.50 W (spec) 1.51 FPS/W	D Older data (2023)	Vitis AI 3.0 Model Zoo performance table

Best published result per workload, across model sizes. For like-for-like groups, open the per-workload tables below.

Show data — per-workload tables

object detection

Like-for-like group	Best result	Model · precision · stack	Grade	Power	Source · date	Rank in group
nano (YOLO-n) · b=1	33.60 FPS	YOLOv8n-Ghost (Hard-Swish, xView3-SAR) INT8 (Vitis AI quantizer, power-of-two) · Vitis AI 3.5 VART, DPU inference only, 2 threads	C	7.50 W spec	arXiv 2507.04842v2 (SAR vessel detection on Kria KV260), Table 6 2025-11-17 (published)	26 of 27
small (YOLO-s) · b=1	12.20 FPS	YOLOv8s-Ghost-P2 (Hard-Swish, xView3-SAR) INT8 (Vitis AI quantizer, power-of-two) · Vitis AI 3.5 VART, DPU inference only, 2 threads	C	7.50 W spec	arXiv 2507.04842v2 (SAR vessel detection on Kria KV260), Table 6 2025-11-17 (published)	22 of 22
small (YOLO-s) · b=unknown	14.79 FPS	pt_yolov5s6_coco_1280_1280_17G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	4 of 4
medium (YOLO-m) · b=unknown	20.12 FPS	pt_yolov6m_coco_640_640_82.4G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	6 of 6
mid-size CNN · b=unknown	5.41 FPS	tf_ssdresnet50v1_fpn_coco_640_640_178.4G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	27 of 34
mobile-class CNN · b=unknown	174 FPS	tf_ssdmobilenetv1_coco_300_300_2.47G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	6 of 6
unknown · b=unknown	276 FPS	pt_yolox-nano_coco_416_416_1G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	18 of 37

instance segmentation

◌

No published instance segmentation results for AMD Kria KV260 Vision AI Starter Kit (K26 SOM) yet.
We only show cited public figures — nothing is estimated to fill gaps. Requests tell us what to source or test next.

pose estimation

Like-for-like group	Best result	Model · precision · stack	Grade	Power	Source · date	Rank in group
mobile-class CNN · b=unknown	352 FPS	pt_movenet_coco_192_192_0.5G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	6 of 6
unknown · b=unknown	20.54 FPS	tf_superpoint_mixed_480_640_52.4G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	5 of 5

image classification

Like-for-like group	Best result	Model · precision · stack	Grade	Power	Source · date	Rank in group
mid-size CNN · b=unknown	214 FPS	pt_OFA-resnet50_imagenet_160_160_0.88_1.8G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	35 of 41
mobile-class CNN · b=unknown	2432 FPS	tf_mobilenetv1_0.25_imagenet_128_128_27M INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	17 of 31
unknown · b=unknown	760 FPS	pt_squeezenet_imagenet_224_224_703.5M INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	6 of 6

semantic segmentation

Like-for-like group	Best result	Model · precision · stack	Grade	Power	Source · date	Rank in group
mobile-class CNN · b=unknown	32.90 FPS	pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_10G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	4 of 4
mid-size CNN · b=unknown	183 FPS	tf2_2d-unet_nuclei_128_128_5.31G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	6 of 6
unknown · b=unknown	73.86 FPS	pt_CFLOW_LIDC_128_128_10.42G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	5 of 6

speech-to-text

◌

No published speech-to-text results for AMD Kria KV260 Vision AI Starter Kit (K26 SOM) yet.
We only show cited public figures — nothing is estimated to fill gaps. Requests tell us what to source or test next.

LLM generation

◌

No published LLM generation results for AMD Kria KV260 Vision AI Starter Kit (K26 SOM) yet.
We only show cited public figures — nothing is estimated to fill gaps. Requests tell us what to source or test next.

depth estimation

Like-for-like group	Best result	Model · precision · stack	Grade	Power	Source · date	Rank in group
unknown · b=unknown	4.51 FPS	pt_fadnetv2_sceneflow_576_960_0.51_201G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	5 of 5

super-resolution

Like-for-like group	Best result	Model · precision · stack	Grade	Power	Source · date	Rank in group
mid-size CNN · b=unknown	77.69 FPS	pt_DRUNet_Kvasir_528_608_0.4G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	4 of 4
unknown · b=unknown	105 FPS	pt_SESR-S_DIV2K_360_640_10.2G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	6 of 6

keypoint / feature detection

Like-for-like group	Best result	Model · precision · stack	Grade	Power	Source · date	Rank in group
unknown · b=unknown	11.31 FPS	tf_HFNet_mixed_960_960_20.09G INT8 · Vitis AI 3.0 (VART)	D	7.50 W spec	Vitis AI 3.0 Model Zoo performance table Older data (2023)	5 of 5

Compare with another platform (107)

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.