Skip to content- Sources: primary, discussion
- Summary: The post asks why a cheap open-weight DeepSeek release drew little reaction compared with earlier ones. The thread answers with deployment arithmetic rather than model quality.
- Why it matters: Reaction to an open-weight release tracks the hardware needed to serve it rather than its licence or its listed price, so a cheap model can still be unreachable for self-hosting.
- Comments: HN commenters put full-precision serving at about 1,664 GB of VRAM and INT4 at about 416 GB, which keeps the model out of reach for self-hosting on commodity hardware.