dgxspark

Browser-based model and inference management for NVIDIA DGX Spark - inventory local and Hugging Face models, manage Ollama and LiteLLM, generate Docker Compose deployments for vLLM, SGLang, llama.cpp, LocalAI, and ComfyUI, with multi-user access, diagnostics, and multi-node support.

Qwen3.8-27B on one DGX Spark GB10: DFlash 2 for 65 tok/s C1 and 120 C4; EAGLE 3/1/4 for ~191 C8, ~330 C16 and ~535 C32; native 262K, pinned SGLang.

GLM-5.2 QuantTrio (Unpruned ) on 4xDGX Spark (GB10): 327K-655K context or up to 5 concurrent agents, one receipe. TP4 + DCP + MTP

A plug and play vLLM manager for DGX Spark. Automatically keeps your engine updated to the newest vLLM release and features out ofthe box multi model launching, memory management, integrated open web ui setup, and more.

submission for Gradient HQ / Parallax DGX Spark Giveaway

Unofficial macOS menu bar monitor for NVIDIA DGX Spark, streaming local DGX Dashboard GPU and memory telemetry.

Blender 4.2.0 compiled for GB10, with CUDA

面向两台 NVIDIA DGX Spark 组成的 vLLM 推理集群的轻量级监控系统。

A minimal CLI for monitoring DGX Spark resource usage. Displays only the essential metrics, simply.

Files for my hashnode and Medium blogs about DGX Spark

Monitor NVIDIA DGX telemetry on macOS with a menu bar status item for RAM and GPU metrics.

Serve Qwen3.8-27B on a single DGX Spark with SGLang, NVFP4 weights, and DFlash2 speculative decoding for up to 387 tok/s aggregate throughput.