minimax-m3.1POST

MiniMax M3.1 Flash runs on 4x DGX Spark via vLLM TP4 at ~40 tok/s

Local deployment report: the 428B/23B-active M3.1 Flash runs on 4x DGX Spark with a day-0 vLLM TP4 recipe at ~40 tok/s prose and ~62 tok/s structured at c1, 262K context self-reported, with 2-Spark EXL3 and pruned NVFP4 options coming.

𝙒𝙞𝙡𝙘𝙤 𝘷𝘢𝘯 𝘽𝙤𝙭𝙩𝙚𝙡 ⚡🚀🧲🌞

@nuonrgArchived 4 days agoOriginally posted 2026.10.23

𝙒𝙞𝙡𝙘𝙤 𝘷𝘢𝘯 𝘽𝙤𝙭𝙩𝙚𝙡 ⚡🚀🧲🌞 @nuonrgOriginally posted 2026.10.23

Open original on X

MiniMax-M3.1-Flash (428B/23B active) is already running on 4x DGX Spark: day-0 vLLM TP4 recipe by @majewskizby, ~40 tok/s prose / ~62 structured at c1, 262K ctx (self-reported). 2-Spark options are coming too: @MiaAI_lab EXL3 3.06bpw (~171GB) + TensorFold recipe soon, and a ~163GB expert-pruned NVFP4. #DGXSpark #LocalAI

x.com/nuonrg/status/2113539212028870814Archived 2026.10.24 12:11