October 18, 2026· ~10 min readLLM Inference Kernels: The Open-Source Arms Race The cost of running a large language model at scale is not the model — it is the inference. A 70-billion-parameter model in BF16 occupies 140 GB of w
August 19, 2026· ~11 min readDeepSeek V4 Flash 0731: Initial Experiences on 2x DGX Spark ## Overview DeepSeek AI released **DeepSeek-V4-Flash-0731**, an updated checkpoint of their mixture-of-experts (MoE) model that power their flaghshi
August 1, 2026· ~14 min readSovereign AI on Your Desk: Why Two DGX Sparks Beat the Cloud In 2024, running a 685-billion-parameter model locally meant a rack of A100s, a six-figure power bill, and a PhD in distributed systems. In 2026, it
May 31, 2026· ~9 min readTwo-Node DGX Spark Cluster: Running DeepSeek V4 Flash at 20 TPS Two NVIDIA DGX Sparks. One 685B-parameter MoE model. A 3× throughput improvement without a single crash. Here's how. ## The Setup: Two GB10 Workstat
May 28, 2026· ~6 min readTaming OOM on DGX Spark: Debugging Unified Memory Pressure in a 2-Node vLLM Cluster Your cluster loads the model. Weights land in GPU memory. Then — crash. `NV_ERR_NO_MEMORY`. The OOM killer takes down `VLLM::Worker_TP`. You restart.
April 25, 2026· ~6 min readDeepSeek V4: 1.6T Parameters, FP4 Precision, and the Huawei NPU Question DeepSeek V4 arrives with two models: a 1.6-trillion-parameter Pro variant and a 284-billion-parameter Flash variant. The headline number is the Pro m