Measured llama.cpp tuning for Flash-Next on a 16GB GPU

Adds NVIDIA's NVFP4 quantisation to Strata so a single 12GB+ RTX 20 to 50-series card can host the MoE.

Dual-3090 build with 128GB RAM reporting about 4,200 tok/s prefill and 111 tok/s decode on long prompts.

Open-source inference engine with an OpenAI/Anthropic-style local API that fits the 125B MoE onto a 12GB consumer GPU plus system RAM. The project behind the model's local-running surge.

Step-by-step recipe to squeeze the 125B model onto one DGX Spark instead of two, with config and measured throughput.