A brief clip of David Ondrej showing a two-unit NVIDIA DGX Spark setup he claims runs DeepSeek V4 Flash at about 95 tokens per second, for viewers curious about small-scale local hosting of open mode…
Continue to AI University →