dgx-spark-qwen38
GitHub

dgx-spark-qwen38 / Models / Kolibri-1

Run Kolibri-1 on a DGX Spark

German and English, by Aleph Alpha. FP8, on vLLM.

Aleph Alpha's model for German and English, in its official FP8 weights on vLLM. It needs about 102 GB of disk, installed on demand. On an ASUS Ascent GX10 with 128 GB: 48.0 tok/s, one request.

Measured

48.0tok/s

one request

  • 121.9 tok/s at 4 streams
  • 165.8 at 8, 202.0 at 16
  • Aleph Alpha's FP8 weights
Measured on
an ASUS Ascent GX10 with 128 GB, the project's reference box
How
the Kolibri lane campaign
Source
docs/kolibri-lane.md

Load it

From the cockpit: open the Lanes view, pick Kolibri-1 and press Load (about 10 minutes to answer). From a terminal, in the repo folder on the box:

./prepare-lane.sh kolibri   # the first time only, builds the lane
./switch-model.sh kolibri   # 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 Kolibri-1 on a DGX Spark

Can a DGX Spark run Kolibri-1?

Yes. Aleph Alpha's model for German and English, in its official FP8 weights on vLLM. It needs about 102 GB of disk, installed on demand. The installer of dgx-spark-qwen38 sets it up, and the cockpit loads it with one button.

How fast is Kolibri-1 on a DGX Spark?

48.0 tok/s, one request, measured on an ASUS Ascent GX10 with 128 GB. 121.9 tok/s at 4 streams. 165.8 at 8, 202.0 at 16. Aleph Alpha's FP8 weights.

Does Kolibri-1 run on a 64 GB DGX Spark?

No: it needs a 128 GB DGX Spark.

The other lanes