macOS menubar app for fast local DeepSeek V4 Pro/Flash, with 1M context.
-
Updated
Aug 16, 2026 - Swift
#
dwarfstar
Here are 5 public repositories matching this topic...
AI-OS-1: model-aware distributed inference OS for bandwidth-bound MoE clusters. Research corpus (paper, forum, frozen schemas). First lab: DeepSeek V4 Flash on DGX Spark + DwarfStar.
-
Updated
Aug 14, 2026 - Python
Running DeepSeek V4 Flash on one NVIDIA DGX Spark (GB10): full 256K context, speculative decode, llama-swap + OpenWebUI wiring, and a thinking-mode tool-call fix. Built on the DwarfStar/ds4 engine.
cuda local-llm llm-inference speculative-decoding deepseek openwebui gb10 llama-swap dgx-spark deepseek-v4 dwarfstar
-
Updated
Jul 28, 2026
Launch DeepSeek-V4-Flash on NVIDIA DGX Spark with optimized DwarfStar 4 scripts for a fast OpenAI-compatible API on port 8888.
cuda self-hosted vllm local-llm llm-inference speculative-decoding deepseek openwebui flashinfer gb10 llama-swap dgx-spark deepseek-v4 dwarfstar
-
Updated
Aug 16, 2026 - Shell
Improve this page
Add a description, image, and links to the dwarfstar topic page so that developers can more easily learn about it.
Add this topic to your repo
To associate your repository with the dwarfstar topic, visit your repo's landing page and select "manage topics."