dgx-spark-qwen38
GitHub

dgx-spark-qwen38 / Models / Nemotron 3.5 Super VL

Run Nemotron 3.5 Super VL on a DGX Spark

A 120B with vision and a 1M-token window, by NVIDIA. With its drafter, on vLLM.

NVIDIA's 120B with vision and a window of a million tokens, on vLLM with its drafter. On an ASUS Ascent GX10 with 128 GB: 69.5 tok/s, rewriting code, one request, reasoning off.

Measured

69.5tok/s

rewriting code, one request, reasoning off

  • 81.7 tok/s in all at 4 streams
  • A window of a million tokens
  • Reads images and video
Measured on
an ASUS Ascent GX10 with 128 GB, the project's reference box
How
the Super VL lane campaign
Source
docs/nemotron-super-lane.md

Load it

From the cockpit: open the Lanes view, pick Nemotron 3.5 Super VL and press Load (about 9 to 11 minutes to answer). From a terminal, in the repo folder on the box:

./prepare-lane.sh nemotron-super   # the first time only, builds the lane
./switch-model.sh nemotron-super   # switches the boot, then prints the commands to run

switch-model.sh never restarts an engine itself: it prints the exact stop and start commands. The cockpit's Load runs them for you.

On a 64 GB DGX Spark

No: it needs a 128 GB DGX Spark. A 64 GB box serves Nemotron 3.5 Bolt and Qwen3.8-27B: what runs on 64 GB.

Use it from your tools

The box serves OpenAI- and Anthropic-compatible APIs on port 30001, with an API key. Claude Code, opencode, Open WebUI and the OpenAI and Anthropic SDKs point there instead of a cloud: see Claude Code with a local model and opencode with a local model.

Install it

One command, on stock DGX OS. The same command updates it later.

curl -fsSL https://raw.githubusercontent.com/hasso5703/dgx-spark-qwen38/main/get.sh | bash

Questions about Nemotron 3.5 Super VL on a DGX Spark

Can a DGX Spark run Nemotron 3.5 Super VL?

Yes. NVIDIA's 120B with vision and a window of a million tokens, on vLLM with its drafter. The installer of dgx-spark-qwen38 sets it up, and the cockpit loads it with one button.

How fast is Nemotron 3.5 Super VL on a DGX Spark?

69.5 tok/s, rewriting code, one request, reasoning off, measured on an ASUS Ascent GX10 with 128 GB. 81.7 tok/s in all at 4 streams. A window of a million tokens. Reads images and video.

Does Nemotron 3.5 Super VL run on a 64 GB DGX Spark?

No: it needs a 128 GB DGX Spark.

The other lanes