Setup guide
Jetson Orin Nano Super for local LLMs.
Same board, a free software update, and a lot more speed. Here is what Super mode changes, what to buy, and how to start.
Updated 2026-10-17. Every speed links to its source; every price is sourced and dated.
What Super mode changes.
| Model (MLC, INT4) | Original tok/s | Super tok/s | Change |
|---|---|---|---|
| SmolLM2 1.7B | 41 | 64.5 | +57% |
| Llama 3.2 3B | 27.7 | 43.1 | +56% |
| Phi-3.5 3.8B | 24.7 | 38.1 | +54% |
| Gemma 2 2B | 21.5 | 35 | +63% |
| Qwen2.5 7B | 14.2 | 21.8 | +54% |
| Llama 3.1 8B | 14 | 19.1 | +36% |
| Gemma 2 9B | 7.2 | 9.2 | +28% |
Same hardware, same source: NVIDIA Technical Blog, JetPack 6.2 Super Mode ↗. The % change is simple arithmetic on the two sourced numbers.
The runtime matters as much as the board: NVIDIA measured 43.1 tok/s on Llama 3.2 3B with MLC (source ↗), while an independent Ollama test measured 6.31 tok/s on the same model (source ↗).
What to buy.
Good to know: NVIDIA ships the dev kit with its 19V power supply, so you do not need to buy one. Storage is not included: NVIDIA asks for a 64GB+ UHS-1 microSD card or an NVMe SSD. source ↗
- Samsung PRO Plus 128GB microSDXC$50.99
Storage (pick one)
A2 / U3 card with room for several quantized models.
$50.99 at Samsung US (list price; out of stock when checked), price at time of writing (2026-10-17) · price source ↗Buy on Amazon - SanDisk Extreme 128GB microSDXC
Storage (pick one)
Fast A2 card for the OS and model files.
Buy on Amazon - Crucial P3 Plus 500GB M.2 2280 NVMe SSD
Optional: NVMe SSD (M.2 2280 slot)
Standard 2280 NVMe drive: faster model loads than microSD and room for bigger models.
Buy on Amazon
Set it up.
- Flash JetPack to a 64GB+ microSD card (or an NVMe SSD) using NVIDIA's quick start guide. Super mode arrived with JetPack 6.2. NVIDIA quick start ↗
- Switch to the Super power mode (MAXN SUPER) described in NVIDIA's announcement. NVIDIA blog ↗
- Pick a runtime. NVIDIA's fastest numbers use MLC. Ollama is the easiest start and its Linux script supports Jetson:
Ollama Linux install ↗curl -fsSL https://ollama.com/install.sh | sh - Measure and share.
ollama run llama3.2:3b --verboseprints the eval rate. Submit your run