DwarfStar hits 3000 t/s using RAM+VRAM hybrid setup
antirez · x · 2026-08-18
Developer demoed running DeepSeek v4 PRO on DGX Station via DwarfStar orchestration. Despite the model exceeding VRAM, leveraging a peculiar RAM+VRAM hybrid setup achieved a crazy prefill speed of 3000 tokens/s. The author emphasizes needing specific inference engines to exploit hardware for local AI.
Related event: DeepSeek v4 local deployment hits 3000 t/s prefill on DGX Station(2 posts)→
More from Infra
- Three CoreWeave co-founders have sold $3.3B in shares since IPO amid credit worries — SumitGup · 2026-10-16
- Cherry pi on a Raspberry Pi: the tiny-model rumor developers are watching — sull · 2026-10-16
- CoreWeave's debt costs falling: DDTL spreads tightened from S+9.62% to S+2.25% — BenBajarin · 2026-10-16
- SoftBank lines up US IPOs for SB Energy and AI robotics spinoff Roze, valuations top $50B — nordicinst · 2026-10-16
- TRL reproduces Modal's mixed-precision async RL: plain GRPO collapses, SC+MIS holds 0.72 — QGallouedec · 2026-10-16
- Hugging Face scales Indic pretraining with 2.8T tokens and an open data pipeline — Hugging Face Blog · 2026-10-16