NVIDIA’s hands-on guides live at build.nvidia.com/spark and their source is on GitHub at NVIDIA/dgx-spark-playbooks. That site is the regularly-updated single source of truth; this page is a grouped index of what was available as of mid-2026.
| Playbook | Time | Link |
|---|
| Set Up Local Network Access | 5 min | link |
| Open WebUI with Ollama | 15 min | link |
| DGX Dashboard + JupyterLab | 30 min | link |
| VS Code | 5 min | link |
| CLI Coding Agent | 20 min | link |
| Playbook | Link |
|---|
| llama.cpp (CUDA, OpenAI-compatible API) | link |
| Nemotron-3-Nano 30B with llama.cpp | link |
| vLLM for Inference | link |
| TensorRT-LLM for Inference | link |
| SGLang for Inference | link |
| LM Studio | link |
| NIM on Spark | link |
| Speculative Decoding | link |
| Playbook | Link |
|---|
| Fine-tune with PyTorch | link |
| Fine-tune with NeMo | link |
| LLaMA Factory | link |
| Unsloth | link |
| FLUX.1 Dreambooth LoRA | link |
| Playbook | Link |
|---|
| NemoClaw with a Local LLM | link |
| Hermes Agent (Nous Research) | link |
| Comfy UI | link |
| Playbook | Link |
|---|
| Single-cell RNA Sequencing | link |
| Portfolio Optimization (cuOpt/cuML) | link |
| CUDA-X Data Science (cuML/cuDF) | link |
| Text to Knowledge Graph | link |
| Optimized JAX | link |
| cuTile Kernels | link |
| Playbook | Link |
|---|
| Connect Two Sparks | link |
See also the connect two Sparks how-to for a condensed walkthrough.
