The short version
DwarfStar is a local DeepSeek V4 Flash/PRO inference engine — a commenter said it was enough to move from one DGX Spark to two.
From the HackerLinks archive
DwarfStar is a local DeepSeek V4 Flash/PRO inference engine — a commenter said it was enough to move from one DGX Spark to two.
The short version
DwarfStar is a local DeepSeek V4 Flash/PRO inference engine — a commenter said it was enough to move from one DGX Spark to two.
Why it caught our attention
Clear praise for a concrete local-AI tool in a hardware-constrained thread.
Where it surfaced on Hacker News
Editorial paraphrase
A commenter said they started with DwarfStar on one Spark, then jumped to two Sparks, and were excited about the future of local LLMs.
Original threadTwo Qwen3 models on one DGX Spark: the residency math