macOS
Homebrew已验证 · 100%
brew install mlxlocal Homebrew formula metadata
安装
其他安装命令
概览
软件包摘要
Array framework for Apple silicon
历史
项目历史与用法
MLX is Apple's open source array framework for machine learning on Apple silicon. The project is published under the ml-explore organization and is described by Apple and the upstream README as an array framework optimized for Apple silicon and its unified memory architecture. The MLX examples citation credits the initial software suite to Awni Hannun, Jagrit Digani, Angelos Katharopoulos, and Ronan Collobert, with a 2023 citation entry.
项目历史
Technically, MLX sits in the NumPy, PyTorch, JAX, and ArrayFire family rather than being only a model runner. Its README emphasizes familiar NumPy-like APIs, C, C++, and Swift bindings, PyTorch-like higher-level neural-network and optimizer packages, composable transformations for automatic differentiation and vectorization, lazy computation, dynamic graph construction, CPU/GPU execution, and unified memory so arrays can be operated on across supported devices without explicit transfers.
使用方式
The framework quickly became the base layer for a cluster of Apple-silicon ML tools: MLX LM for language models, MLX examples for reference implementations across text, image, audio, video, and multimodal models, and community-converted model weights on Hugging Face. LM Studio also shipped an MLX engine for on-device LLM use on Apple-silicon Macs. In package managers, the `mlx` package is the foundation dependency; users install higher-level packages such as `mlx-lm` when they want complete model workflows.
来源
安全态势
风险级别：绿色
narrow executable package without higher-risk signals.
风险分类器
绿色 风险 · 低 置信度 · appliance
原因
- narrow executable package without higher-risk signals
信号
- metadata:no-higher-risk-signals
安装行为
- formula 元数据中未记录 Homebrew post-install 钩子。
- Homebrew bottle 元数据适用于 3 个平台目标。
- 安装时包含 1 个运行时依赖。
- 构建元数据列出 6 个构建依赖。
建议审查
在无人值守的代理使用前，请检查该工具是否读取明文凭据、写入远程状态、发布制品或调用插件。
可执行文件
已安装的可执行文件
| 命令 | 类型 | 暴露范围 | 备注 |
|---|---|---|---|
mlx.distributed_config | cli | 全局可执行文件 | |
mlx.launch | cli | 全局可执行文件 |
新鲜度
版本和新鲜度
这些信号区分页生成时间、软件包管理器活动和上游发布比较。只有存在证据 URL 和可比较版本时，才会提示版本落后。
页面生成时间2026-07-25
管理器版本0.32.0
管理器更新时间2026-07-08
本地数据OK
上游当前
检测到的最新版本v0.32.0
https://github.com/ml-explore/mlx
- OK没有生成新鲜度警告。
安装元数据
软件包元数据
| 软件包键 | brew:mlx |
|---|---|
| 版本 | 0.32.0 |
| 软件包管理器 | Homebrew |
| 软件包管理器页面 | https://formulae.brew.sh/formula/mlx |
| 主页 | https://ml-explore.github.io/mlx/build/html/index.html |
| 仓库 | https://github.com/ml-explore/mlx |
| 上游文档 | https://ml-explore.github.io/mlx/build/html/index.html |
| 许可证 | MIT AND Apache-2.0 |
| 源码归档 | https://github.com/ml-explore/mlx/archive/refs/tags/v0.32.0.tar.gz |
| 最后更新 | 2026-07-08T14:26:04Z |
| Pulse | updated |
| 依赖 | python@3.14 |
| 构建依赖 | cmake, fmt, nanobind, nlohmann-json, python-setuptools, robin-map |
| Bottle | 可用 (于 arm64_sequoia, arm64_sonoma, arm64_tahoe) |
| Homebrew post-install | 未定义 |
| 服务 | 未声明 |
注册表事实
源数据库详情
| Source Database | Homebrew formula API |
|---|---|
| Tap | homebrew/core |
| Full Name | mlx |
| Version Scheme | 0 |
| Revision | 0 |
| Head Version | HEAD |
| Requirements |
|
| Bottle Stable Root URL | https://ghcr.io/v2/homebrew/core |
| Deprecated | no |
| Disabled | no |
| Keg Only | no |
| URL Keys |
|
源数据库匹配
其他软件包管理器记录
匹配项来自外部软件包管理器索引，并与本地 Automic Vault 软件包链接分开显示。
MacPorts95%
mlx
sudo port install mlx- normalized package name match
- 匹配方式：Mlx
来源线索
由仓库数据生成
此页面由 av-web 从 scripts/generate-pkg-sqlite.py 生成的私有软件包 SQLite 工件提供。
Combined YAML source
View the package source record on GitHub.
使用的来源
- Geiger risk classifier
- Nucleus package database
- av.db category and tag curation
- cross-ecosystem install command graph
- curated package history
- external package-manager database matches
- package relationship graph
- package version freshness
- package-page enrichment