Tiny DeepSeek 1.5B Models Run on $249 NVIDIA Jetson Nano

Youtuber, Ominous Industries, ran a couple of versions of the DeepSeek R1 1.5B of models running locally on the NVIDIA Jetson Nan. The newly released distilled DeepSeek models were explroed. The DeepSeek R1 1.5B model delivers impressive performance with plenty of room to spare on the Jetson.

He shows the installation process, followed by a series of tests, including a Python reasoning test. He compared the DeepSeek R1 distilled 1.5B models to the Mini Llama 3.1 1B model. He set up and tested the FP16 version of the DeepSeek R1 Distilled 1.5B Qwen model, running seamlessly in WebUI with an Ollama backend.

Here are the specifications and cost for the NVIDIA Jetson Nano:

Specifications:

GPU: 128-core NVIDIA Maxwell architecture GPU

CPU: Quad-core ARM Cortex-A57 MPCore processor

Memory:

Jetson Nano 4GB Developer Kit: 4GB 64-bit LPDDR4, 25.6 GB/s, 1600MHz

Jetson Nano 2GB Developer Kit: 2GB 64-bit LPDDR4, 25.6 GB/s, 1600MHz (for the 2GB version)

Storage: MicroSD card slot for OS and storage

Video:

1x HDMI 2.0, 1x DisplayPort 1.3 over USB-C

Supports resolutions up to 4K 60Hz

Camera: 1x MIPI CSI-2 D-PHY interface, supports 2 cameras (in newer versions like V3)

Connectivity:

Gigabit Ethernet, 4x USB 3.0 ports

40-pin expansion header with GPIO pins

Power:

5W to 10W power consumption; can be powered via micro-USB (4GB version) or USB-C (2GB version)

Performance: Up to 472 GFLOPS of compute performance (FP16)

Software: Supported by NVIDIA JetPack SDK, which includes CUDA, cuDNN, and TensorRT for AI and machine learning applications.

There is a range of pricing for the Nvidia Jetson Nano and Nvidia Orin Nano.