Validated Models

Explore a curated collection of AI models that have been validated and optimized for BOS Eagle-N hardware. This includes large language models (LLMs), vision models, and other neural networks with verified performance benchmarks and deployment guides.

---

• LLM-VLM Models
• Vision Models

Validated AI Models and Their Performance

Performance metrics for validated AI models running on BOS Eagle-N hardware. Includes measured throughput, latency, and optimization details for large language models, vision models, and object detection networks tested on the platform.

LLM and VLM

Model	Batch size	ttft (time to first token)	t/s/u	target t/s/u	t/s	DRAM / PCIe
QWEN 2.5VL 3B	2	1183.11 ms	9.51	48.57	19.01	LPDDR5x 7.5GHz / Gen5
QWEN 2.5VL 7B	2	1422.01 ms	4.36	22.18	8.71	LPDDR5x 7.5GHz / Gen5
LLaMA 3.1 8B	16	232.63 ms	4.82	29.76	77.17	LPDDR5x 7.5GHz / Gen5
LLaMA 3.2 1B	16	51.4 ms	22.71	92.15	363.36	LPDDR5x 7.5GHz / Gen5
LLaMA 3.2 3B	16	100.3 ms	9.97	53.34	159.53	LPDDR5x 7.5GHz / Gen5

Object detection

Model	Batch size	Input size	FPS	Target FPS	DRAM / PCIe
SSR	1	[ 640 x 384 ] x 6	4.61	21.7	LPDDR5x 7.5GHz / Gen5
YOLOv8s	1	256 x 256	164	1632.08	LPDDR5x 7.5GHz / Gen5
	1	320 x 320	138	1415.09	LPDDR5x 7.5GHz / Gen5
DeepLab V3+	1	512 x 1024	18.03	173.65	LPDDR5x 7.5GHz / Gen5
PDL	1	512 x 1024	12.24	173.65	LPDDR5x 7.5GHz / Gen5
FastOFT	1	384 x 1280	16.13	141.60	LPDDR5x 7.5GHz / Gen5
ResNet50	4	224 x 224	976	8969	LPDDR5x 7.5GHz / Gen5