DeepSeek-V4-Flash benchmark: MacBook scores 54% vs 52% over 2× DGX Spark

In a community 89-task Terminal-Bench 2.1 run, an 80.8 GiB GGUF on a 128 GB M5 Max MacBook scored 54%, beating 2× DGX Spark at 52%. The DGX setup ran the native mixed FP8/FP4 checkpoint with DSpark speculative decoding.