DeepSeek V4.1 Flash on 4x DGX Spark: my numbers vs Patrick Moorhead's results via Wccftech.
A couple of weeks ago I read this Wccftech article about Patrick Moorhead's 4 Spark setup.
Proof1 community thread
Key points
- I meant to post this a few days ago but ran some more tests first, and then my workstation had a problem after a kernel and NVIDIA driver update.
- Initially I thought this Patrick Moorhead guy ran these benchmarks and was interested in the setup.
- I'm just severely disappointed over the lack of info, and over the misreported memory requirement in the Wccftech article.
- DeepSeek-V4.1-Flash doesn't fit comfortably on 4x Sparks.
Sources (1)
- [1]DeepSeek V4.1 Flash on 4x DGX Spark: my numbers vs Patrick Moorhead's results via Wccftech.r/LocalLLaMA (top, daily) · Oct 16, 07:11 AM
A couple of weeks ago I read this Wccftech article about Patrick Moorhead's 4 Spark setup.
I meant to post this a few days ago but ran some more tests first, and then my workstation had a problem after a kernel and NVIDIA driver update.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 16, 2026Welcome Beam: a 501B open model you can own, customize, and scale
- Oct 15, 2026unslothai/unsloth v0.1.908-beta: NVIDIA Nemotron 3.5 + Meta Muse Spark Open
- Oct 15, 2026NVIDIA Nemotron 3.5 Super VL Powers Specialized Agents with Multimodal Reasoning
- Oct 15, 2026StepFun Releases Step 5 Preview Open Weights: 600B MoE, 27B Active, 1M Context
- Oct 15, 2026Local and Distributed Large LLM Inference with MLX and RDMA on Mac Studio
- Oct 14, 2026GPT-Image-style photo editing on your own GPU (12 GB+): I built a zero-setup app for Qwen-Image 2.1 on stable-diffusion.cpp
