User runs REAP-pruned DeepSeek-V4-Flash on Ascent GX10

A Reddit user reports running a 162B REAP-pruned NVFP4 DeepSeek-V4-Flash model on a single Ascent GX10 Spark using a patched vllm Docker image. The setup maintains consistent performance at long context according to the user's tests.