alzeo
Alzeo is an open-source installer that turns an NVIDIA DGX Spark, or any GB10 machine (ASUS, Dell, HP, Lenovo, MSI, GIGABYTE, Acer), into a private AI server in
What is alzeo?
Alzeo is an open-source installer that turns an NVIDIA DGX Spark, or any GB10 machine (ASUS, Dell, HP, Lenovo, MSI, GIGABYTE, Acer), into a private AI server in one command: measured model lanes, OpenAI and Anthropic APIs, and a local cockpit.
Overview of alzeo
What it is
dgx-spark-qwen38 is a software repository that provides a single command to turn a DGX Spark or ASUS Ascent GX10 into a private AI server. It installs and pins seven measured AI lanes for text (up to 1M tokens), images, and video with sound. All processing stays on the local machine; no data leaves the box. The system is controlled from a local cockpit accessible via a laptop or phone browser.
Key features
- Pinned and proved: images by digest, checkpoints by revision; the lane that serves proves it serves before the install ends, and an update uses the same command.
- One pool of unified memory with lanes that take turns; the cockpit shows what holds memory and loads a lane with one button.
- Clients that work out of the box: OpenAI- and Anthropic-compatible APIs, opencode wired to the local model, Claude Code, VS Code Copilot, Open WebUI and Cursor.
- Measured performance: Qwen3.8-27B at 71.4 tok/s greedy median with 1M context, Qwen3.8-Flash-Next 176B at 47.9 tok/s on code, Qwen-Image 2.1 at 7.5 s for 1024×1024 with Turbo.
- Includes multiple models: Qwen3.8 family, Kolibri-1, Nemotron 3.5 Bolt, Nemotron 3.5 Super VL, FP4-Local Flash-Next, Ornith 1.5 122B, MiniMax-H3 video with sound, opt-in speech, and a System One endpoint with calibrated probabilities.
- Seven switchable targets for the two Qwen text lanes, with speculation costing no quality.
Use cases
Running a private AI server on a DGX Spark or ASUS Ascent GX10 for text generation, image generation, video generation, and speech tasks without sending data to external services. Suitable for local development, agent workflows, and multi-model serving from a single machine.
Getting started
bashcurl -fsSL https://raw.githubusercontent.com/hasso5703/dgx-spark-qwen38/main/get.sh | bashAlternatively, clone the repository and run ./install.sh.
License and pricing
The license is MIT. The pricing type is open source.
Based on the project's public README and repository metadata.