10T tokens/day of a rumored DeepSeek v4.1 flash on one GB300 NVL72?
_xjdr · x · 2026-10-29
AI inference commentator xjdr asks: what would you do with 10 trillion tokens per day of DeepSeek v4.1 flash served from a single GB300 NVL72 rack? The terse question implicitly teases an unconfirmed "DeepSeek v4.1 flash" version and the very high throughput next-generation NVL72 racks could deliver for serving.
More from Infra
- Why 4x DGX Sparks beat a single 512GB M5 Ultra Mac for local AI workloads — JFPuget · 2026-10-29
- MemTensor's MemT-1.0 report: 66x less storage than full-context KV cache at 128K tokens — jiqizhixin · 2026-10-29
- "Insert coins to continue": meme laments the end of cheap AI tokens — Suspicious_Orchid770 · 2026-10-29
- Skip ComfyUI: run Flux-3-Image-Klein locally with stable-diffusion.cpp in four files — srigi · 2026-10-29
- GPU unlock is pointless for most users: quantized models, decode and gaming see zero speedup — Justify_87 · 2026-10-29
- US AI Power in Three Stages: 10GW+ Nuclear Contracts, Geothermal Live, SMRs Permitting — ingliguori · 2026-10-29