MiniMax M3.1 Flash 428B on two DGX Sparks: EXL3 2.15bpw, 39 tok/s single stream

SAGE EXL3 2.15bpw quantization with TP2 over RoCE on two DGX Sparks gives 300K context, 726K KV pool, up to 39 tok/s single stream and 65–78 tok/s at four users.

Tech2Wild

@Tech2WildArchived 3 days agoOriginally posted 2026.10.24

Tech2Wild @Tech2WildOriginally posted 2026.10.24

Open original on X

🔥 MiniMax M3.1 Flash (428B) on 2 NVIDIA DGX Sparks, TP2 over RoCE 🧠 Quant by @ViC305 (SAGE EXL3 2.15 bpw) 🙌 📏 300K context, 726K KV pool 🏎️ Up to 39 tok/s single stream, 65-78 tok/s at 4 users 🛠️ 2 vLLM-EXL3 patches for TP2 Recipe + benchmarks 👇 https://github.com/tonyd2wild/MiniMax-M3.1-Flash-EXL3-TP2-2x-DGX-Spark

x.com/tech2wild/status/2113893367784276026Archived 2026.10.25 12:07

Open original on X