M3.1 Flash 在 4× DGX Spark 上 TP4 首测：散文解码 39.7 tok/s，262k 上下文
作者在 4 台 DGX Spark 上以 TP4 跑 M3.1 Flash 并给出实测：c1 散文解码 39.7 tok/s、结构化解码 62.2 tok/s，c16 聚合 166/261 tok/s，262k 上下文约 1.37M token FP8 KV 缓存，用 MiniMax 自带 NVFP4 专家与 MXFP
MiniMax-M3.1-Flash TP4 on 4× DGX Spark, day 0 🐇 • Prose decode: 39.7 tok/s at c1 • Structured decode: 62.2 tok/s at c1 • Aggregate: 166 tok/s prose, 261 tok/s structured at c16 • speedbench quick: 38.4 tok/s across 30 prompts • 262k context, ~1.37M-token FP8 KV cache • MiniMax's own NVFP4 experts + MXFP8 weights, bundled DSpark draft • Runs on GB10 (sm_121): FP8 KV, Triton sparse attention, Marlin experts (the NVFP4 KV cache is SM100-only) First measured bring-up. Speed tuning and quality gates are next. Full recipe, measurements and credits: https://github.com/knapcio/MiniMax-M3.1-Flash-4x-DGX-Spark-TP4