deepseek-v4.1POST
MiniMax M3.1 Flash open weights: mixed results vs DeepSeek V4.1 Flash on official table
MiniMax released M3.1 Flash weights (428B total, 23B active, 1M context); its official comparison table shows it behind DeepSeek V4.1 Flash on DeepSWE (73.7 vs 77.0) and Terminal Bench (38.1 vs 42.3) but slightly ahead on Automation-Bench (62.8 vs 60.5) and tool-free HLE (43.4 vs 42.6).
墨无归
@MoblacksoilArchived 9 days agoOriginally posted 2026.10.23
墨无归 @MoblacksoilOriginally posted 2026.10.23
MiniMax 把 M3.1 Flash 的权重放出来了，能自己下载部署。总参数 428B，每次只激活 23B，上下文最长 1M；官方报的智能体成绩有 BankerToolBench 76.2、DeepSWE 73.7%、OSWorld V2 61.7。 模型页上和 DeepSeek V4.1 Flash 同表比，是互有胜负：写代码的 DeepSWE（73.7 对 77.0）和 Terminal Bench（38.1 对 42.3）落后，办公自动化 Automation-Bench（62.8 对 60.5）和无工具 HLE（43.4 对 42.6）略高。
This post quotes another post: “MiniMax M3.1 Flash open weights: 428B total, 23B activated”
x.com/moblacksoil/status/2113453353883607094Archived 2026.10.23 12:05