Meta 发布 Muse Glimmer 技术报告:30B 开源权重模型面向消费级硬件智能体任务
Today we published a technical report on Muse Glimmer, our 30B open-weight model built for agentic tasks on consumer hardware. It runs at practical speeds on a Mac or PC with a single consumer GPU at 4-bit precision, and leads comparably-sized open models on tool use and multi-step planning. Read about the model architecture, distillation recipe, quantization and more: https://t.co/H7n9qeDhxn
Meta 发布 Muse Glimmer 技术报告,这是一个 30B 开源权重模型,面向消费级硬件上的智能体任务。该模型在 Mac 或 PC 上以 4-bit 精度、单张消费级 GPU 即可达到实用速度,并在工具调用和多步规划上领先同尺寸开源模型。报告涵盖模型架构、蒸馏方案和量化等内容。
原文给出 30B 开源权重模型在单张消费级 GPU 上 4-bit 运行的设定,适合围绕工具调用与多步规划做同尺寸开源模型对比实测。
粉丝 87万
认证:@Meta 的关联账号
We're building the future of personal superintelligence to help people create, connect, and achieve more in life.
- 浏览
- 3.6万
- 点赞
- 483
- 评论
- 54
- 分享
- 55
- 收藏
- 177
数字来自平台,为最近一次抓取时的数据。
来源:X · AI at Meta · x.com