Models

Ready-to-use models

Pick a model and deploy quickly. Filter by FPGA target to see what fits your board.

VisionLLM (experimental)NLP

Filter by FPGA:

SSD MobileNet v2

Fast object detection for Kria edge devices.

DPUINT8

Tasks: Object detection

I/O: RGB image (e.g., 300×300) → boxes + classes + scores

Targets: KV260, KR260

Great default for KV260/KR260 vision pipelines.

YOLOv3 COCO 416

Accurate detection with an edge-friendly pipeline.

DPUINT8

Tasks: Object detection

I/O: RGB image 416×416 → boxes + classes + scores

Targets: KV260, KR260

MiniLM Embeddings

Fast embeddings for search/RAG on embedded devices.

CPUFP16

Tasks: Embeddings, Semantic search

I/O: Text → vector embedding

Targets: KV260, KR260, PYNQ-Z2

Tiny LLM (~0.5B) (Experimental)

Small chat model — CPU-only on embedded boards (experimental).

CPUINT4

Tasks: Chat, Q&A

I/O: Text prompt → text output

Targets: KV260, KR260

Shown for experimentation. Expect low tokens/sec on embedded CPUs.

U55 Vision Pack (Batch)

Datacenter-oriented vision pack for high throughput on U55.

DPUINT8

Tasks: Object detection, Classification

I/O: Batch images → detections / logits

Targets: Alveo U55

Need a model not listed?

We can add or optimize it for your FPGA target.

Tell us the model, input shape, and latency/throughput goal.