deepseek-v4.1POST

DeepSeek V4.1 Flash on dual-cable DGX Spark: 93.6 tok/s single user

Adding a second network cable lifts single-user throughput to 93.6 tok/s (1.93x vLLM's 48.5) and 213.8 tok/s at 6 users, with long-prompt reading also faster; code is open sourced.

Tech2Wild

@Tech2WildArchived 12 days agoOriginally posted 2026.10.18

Tech2Wild @Tech2WildOriginally posted 2026.10.18

Open original on X

Bruh ! 🔌 Turned on the 2nd network cable between the Sparks and TensorFold got even faster: ⚡️ 93.6 tok/s for 1 user (vLLM: 48.5), 1.93x 👥 213.8 tok/s with 6 users (+29%) 📖 Long prompt reading now even with vLLM https://github.com/tonyd2wild/DeepSeek-V4.1-Flash-vLLM-DGX-Spark

This post quotes another post: x.com/tech2wild/status/2111658578100809966

x.com/tech2wild/status/2111666374032425064Archived 2026.10.18 16:25