MiniMax 开放 M3.1 Flash 模型权重，总参数 428BChinese only
MiniMax 宣布开放 M3.1 Flash 模型权重；这是一款总参数 428B、激活参数 23B 的模型。公司称其能力较上一代 M3 全面提升，并通过稀疏注意力、量化感知训练等优化推理效率；模型于 10 月 19 日上线，72 小时内真实用户处理量超过 100 万亿 Token。
2 developments3 reports2 sources
Event developments 2
MiniMax releases M3.1 Flash model weights, with 428B total parameters
MiniMax announced the release of M3.1 Flash model weights. The model has 428B total parameters and 23B active parameters. The company says its capabilities improve across the board over M3 and that sparse attention and quantization-aware training improve inference efficiency. After its October 19 launch, real users processed more than 100 trillion tokens within 72 hours.
View source reports · 1 sources · 1 reports
MiniMax releases M3.1 Flash, with model weights planned for later
MiniMax has released M3.1 Flash, outlining its capabilities for complex agent tasks, software development and computer use alongside inference speed and pricing. The digest distinguishes the vendor’s benchmark results and services already online from its plan to make the weights available later.
View source reports · 2 sources · 2 reports
- ginobefun (@hongming731) · MiniMax releases M3.1 Flash, with model weights planned for later
- IT之家 科技新闻 · MiniMax launches M3.1 Flash, claims nearly 3× faster inference
All reports 3
MiniMax releases M3.1 Flash model weights, with 428B total parameters
MiniMax announced the release of M3.1 Flash model weights. The model has 428B total parameters and 23B active parameters. The company says its capabilities improve across the board over M3 and that sparse attention and quantization-aware training improve inference efficiency. After its October 19 launch, real users processed more than 100 trillion tokens within 72 hours.
2 more sources reported this
- IT之家 科技新闻
- ginobefun (@hongming731)
MiniMax launches M3.1 Flash, claims nearly 3× faster inference
MiniMax says M3.1 Flash is available through MiniMax API, MPlan and MiniMax Code. It claims nearly 3× faster inference and 70% lower usage costs; model weights are due to be released later. The supplied report does not specify how those figures were measured.
2 more sources reported this
- IT之家 科技新闻
- ginobefun (@hongming731)
MiniMax releases M3.1 Flash, with model weights planned for later
MiniMax has released M3.1 Flash, outlining its capabilities for complex agent tasks, software development and computer use alongside inference speed and pricing. The digest distinguishes the vendor’s benchmark results and services already online from its plan to make the weights available later.
1 more source reported this
- IT之家 科技新闻