minimax-m3.1POST
MiniMax M3.1 Flash: self-hosting 428B weights vs official API pricing
Advice on choosing between self-hosting the 428B/23B-active open weights (needs multi-GPU vLLM/SGLang, DeepSWE 73.7%) and the official API at ¥1 input / ¥4 output per million tokens, after estimating monthly token use.
daviswu
@DaviswhArchived 4 days agoOriginally posted 2026.10.23
daviswu @DaviswhOriginally posted 2026.10.23
早上那条权重公开了，怎么选更实在： 自托管：HF MiniMaxAI/MiniMax-M3.1-Flash，428B 总参 / 23B 激活，DeepSWE 73.7%。有多卡、能跑 vLLM/SGLang 的再下；单卡别硬上，先等官方 8-bit / 社区量化。 官方 API：国内按量输入 ¥1、输出 ¥4 / 百万 tokens。要稳定接口、不想养卡的直接充值用。 判断：先拿自己的真实 agent/写码任务估一下月 token，再决定养卡还是走 API；别被 428B 唬住。
This post quotes another post: “MiniMax M3.1 Flash weights open: 428B total, 23B active, DeepSWE 73.7%”
x.com/daviswh/status/2113578086637707615Archived 2026.10.24 12:11