# NVIDIA DGX Spark

> DGX Spark (GB10 Grace Blackwell, 128GB unified memory) for local AI: setup, DGX OS, unified memory & ARM64, llama.cpp/Ollama/vLLM/LM Studio/ComfyUI on Spark, agents, clustering, and a cross-project troubleshooting casebook.
> Covers: NVIDIA DGX Spark as a local-AI machine, verified against DGX OS 7.5.0 (2026-07-18): hardware & the GB10 unified-memory model, first boot & setup, DGX OS & updates/recovery, containers & NGC, ARM64/aarch64 realities, running llama.cpp / Ollama / vLLM / LM Studio / ComfyUI on Spark, agents on Spark (incl. the Hermes playbook), multi-Spark clustering & benchmarking, fine-tuning & NVFP4 quantization, the full official playbooks catalog (66), a cross-project troubleshooting casebook (llama.cpp/Ollama/vLLM issues with hardware provenance), and an inference stack picker.
> Not covered: DGX Station (GB300) and other DGX systems (Station playbooks are catalogued but excluded from Spark guidance); generic per-tool depth covered by the dedicated llama-cpp / vllm / ollama / lmstudio / comfyui wikis; CUDA programming beyond Spark-specific kernel playbooks; purchasing/pricing advice; and Windows/WSL (Spark runs DGX OS).
> Current as of: 2026-07-18 (7.5.0)

- [LLM Wiki](/raw/dgx-spark/README.md)
- [Knowledge Base Index](/raw/dgx-spark/wiki/index.md)
- [Agents on Spark](/raw/dgx-spark/wiki/concepts/agents-on-spark.md)
- [Clustering Multiple DGX Sparks](/raw/dgx-spark/wiki/concepts/clustering.md)
- [ComfyUI and Image Generation on Spark](/raw/dgx-spark/wiki/concepts/comfyui-and-image-gen.md)
- [Containers and NGC](/raw/dgx-spark/wiki/concepts/containers-and-ngc.md)
- [DGX OS and Updates](/raw/dgx-spark/wiki/concepts/dgx-os-and-updates.md)
- [Fine-Tuning and Quantization on Spark](/raw/dgx-spark/wiki/concepts/fine-tuning-and-quantization.md)
- [First Boot and Setup](/raw/dgx-spark/wiki/concepts/first-boot-and-setup.md)
- [llama.cpp on Spark](/raw/dgx-spark/wiki/concepts/llama-cpp-on-spark.md)
- [LM Studio on Spark](/raw/dgx-spark/wiki/concepts/lm-studio-on-spark.md)
- [Ollama on Spark](/raw/dgx-spark/wiki/concepts/ollama-on-spark.md)
- [Overview and Hardware](/raw/dgx-spark/wiki/concepts/overview-and-hardware.md)
- [System Operations](/raw/dgx-spark/wiki/concepts/system-operations.md)
- [Unified Memory and ARM64 on DGX Spark](/raw/dgx-spark/wiki/concepts/unified-memory-and-arm64.md)
- [vLLM on Spark](/raw/dgx-spark/wiki/concepts/vllm-on-spark.md)
- [DGX Spark Playbooks Catalog](/raw/dgx-spark/wiki/entities/playbooks-catalog.md)
- [Activity Log](/raw/dgx-spark/wiki/log.md)
- [DGX Spark Release Notes Digest](/raw/dgx-spark/wiki/summaries/release-digest.md)
- [Stack Picker: Choosing an Inference Stack on Spark](/raw/dgx-spark/wiki/syntheses/stack-picker.md)
- [Troubleshooting Casebook: llama.cpp, Ollama, and vLLM on Spark](/raw/dgx-spark/wiki/syntheses/troubleshooting-casebook.md)

## XL edition (Pro)

> Deeper per-item reference pages. Requires a subscription (/pro) — send
> `Authorization: Bearer <api key>` on the /raw/ URLs below.

- [XL Edition Index](/raw/dgx-spark/wiki-xl/index.md) — Pro
- [Deep Reference — Performance Benchmarking Methodology on DGX Spark](/raw/dgx-spark/wiki-xl/reference/benchmarking.md) — Pro
- [Casebook: llama.cpp on DGX Spark — Per-Issue Deep Dives](/raw/dgx-spark/wiki-xl/reference/casebook-llama-cpp.md) — Pro
- [Casebook: Ollama on DGX Spark — Per-Issue Deep Dives](/raw/dgx-spark/wiki-xl/reference/casebook-ollama.md) — Pro
- [Casebook: vLLM on DGX Spark — Per-Issue Deep Dives](/raw/dgx-spark/wiki-xl/reference/casebook-vllm.md) — Pro
- [CLI Coding Agent — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-cli-coding-agent.md) — Pro
- [Comfy UI — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-comfy-ui.md) — Pro
- [Connect Three Sparks — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-connect-three-sparks.md) — Pro
- [Connect to Your Spark — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-connect-to-your-spark.md) — Pro
- [Connect Two Sparks — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-connect-two-sparks.md) — Pro
- [Playbook Deep Reference — CUDA-X Data Science (cuDF / cuML)](/raw/dgx-spark/wiki-xl/reference/playbook-cuda-x-data-science.md) — Pro
- [Playbook Deep Reference — cuTile Kernels (TileGym)](/raw/dgx-spark/wiki-xl/reference/playbook-cutile-kernels.md) — Pro
- [DGX Dashboard — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-dgx-dashboard.md) — Pro
- [Playbook Deep Reference — FLUX.1 Dreambooth LoRA Fine-Tuning](/raw/dgx-spark/wiki-xl/reference/playbook-flux-finetuning.md) — Pro
- [Hermes Agent — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-hermes-agent.md) — Pro
- [Playbook Deep Reference — Isaac Sim and Isaac Lab](/raw/dgx-spark/wiki-xl/reference/playbook-isaac.md) — Pro
- [Playbook Deep Reference — Optimized JAX on Spark](/raw/dgx-spark/wiki-xl/reference/playbook-jax.md) — Pro
- [Live VLM WebUI — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-live-vlm-webui.md) — Pro
- [llama.cpp on Spark — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-llama-cpp.md) — Pro
- [Playbook Deep Reference — LLaMA Factory](/raw/dgx-spark/wiki-xl/reference/playbook-llama-factory.md) — Pro
- [LM Studio on Spark — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-lm-studio.md) — Pro
- [Multi-Agent Chatbot — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-multi-agent-chatbot.md) — Pro
- [Multi-modal Inference (TensorRT) — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-multi-modal-inference.md) — Pro
- [Connect Multiple Sparks Through a Switch — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-multi-sparks-through-switch.md) — Pro
- [Playbook Deep Reference — NCCL for Two Sparks](/raw/dgx-spark/wiki-xl/reference/playbook-nccl.md) — Pro
- [Playbook Deep Reference — NeMo AutoModel Fine-Tuning (nemo-fine-tune)](/raw/dgx-spark/wiki-xl/reference/playbook-nemo-fine-tune.md) — Pro
- [Playbook Deep Reference — NemoClaw (Local Agent Sandbox + Applications)](/raw/dgx-spark/wiki-xl/reference/playbook-nemoclaw.md) — Pro
- [Playbook Deep Reference — Nemotron Model Family on DGX Spark](/raw/dgx-spark/wiki-xl/reference/playbook-nemotron.md) — Pro
- [NIM on Spark — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-nim-llm.md) — Pro
- [Playbook Deep Reference — NVFP4 Quantization](/raw/dgx-spark/wiki-xl/reference/playbook-nvfp4-quantization.md) — Pro
- [Ollama on Spark — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-ollama.md) — Pro
- [Open WebUI with Ollama — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-open-webui.md) — Pro
- [Playbook Deep Reference — Fine-Tune with PyTorch (single and dual Spark)](/raw/dgx-spark/wiki-xl/reference/playbook-pytorch-fine-tune.md) — Pro
- [Playbook Deep Reference — Unsloth on DGX Spark](/raw/dgx-spark/wiki-xl/reference/playbook-unsloth.md) — Pro
- [vLLM on Spark — Full Playbook Reference](/raw/dgx-spark/wiki-xl/reference/playbook-vllm.md) — Pro