> For the complete documentation index, see [llms.txt](https://docs.modular.com/llms.txt).
> Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

# FAQ

If this page doesn't answer your question, please ask us on our [Modular
forum](https://forum.modular.com) or [Discord
channel](https://www.discord.gg/modular).

## Distribution

### What operating systems do you support? {#system-requirements}

You can install `modular` on Mac and Linux operating systems.

For more details, see the [system
requirements](https://docs.modular.com/max/packages.md#system-requirements).

### What are the GPU requirements? {#gpu-requirements}

The Modular Platform supports both CPUs and GPUs, so you don't always need a
GPU to serve a model—although some larger models do require a GPU.

For details about GPU support, see our [list of compatible
GPUs](https://docs.modular.com/max/packages.md#gpu-compatibility).

### Will MAX be open-sourced?

We want to contribute a lot to open source, but we also want to do it right.
Our team has decades of experience building open-source projects, and we
believe it's very important to create an inclusive and vibrant
community, which takes a lot of work.

We've already begun open-sourcing parts of the MAX framework, including our
[Python serving library](https://github.com/modular/modular/tree/main/max/python/max/serve),
[MAX model architectures](https://github.com/modular/modular/tree/main/max/python/max/pipelines/architectures),
and
[GPU kernels](https://github.com/modular/modular/tree/main/max/kernels/src/nn).

To get the latest updates, [join our
community](https://www.modular.com/community).

## Functionality

### What clouds and services can I deploy MAX onto?

You can deploy our MAX container across a variety of VM and Kubernetes-based
cloud services, including AWS, GCP, and Azure. To get started with any of them,
check out our
[tutorial to deploy a model to the cloud](https://docs.modular.com/max/deploy/local-to-cloud.md).

### Can I run MAX locally?

Yes. MAX has support for MacOS and ARM hardware, meaning it can be run on your
local laptop for exploration and testing purposes.

### Does MAX support distributed inference of large models?

Yes. MAX runs tensor-parallel inference across multiple GPUs using the
distributed layers in `max.nn` and distributed ops in `max.graph`. Production
pipelines such as Llama 3, DeepSeek, and Mistral ship with multi-GPU
implementations, and you can serve them across multiple GPUs by passing
`--devices=gpu:0,1` (or `--devices=gpu:all`) to `max serve`.

### What if my BF16 results differ slightly from PyTorch? {#bf16-precision}

MAX uses FMA (fused multiply-add) contraction when compiling operations within
a single graph. This means a `bfloat16` multiply followed by an add becomes a
single FMA instruction that rounds once instead of twice. The FMA result is
more mathematically accurate, but individual elements can differ from PyTorch
by up to 1 ULP (unit in the last place). This has no impact on model-level
accuracy. For details, see [Numerical
precision](https://docs.modular.com/max/develop/dtypes.md#numerical-precision).

## Installation

### Can I install both stable and nightly builds?

Yes, it's safe and easy to use the stable
and nightly builds for different projects, each with their own virtual
environment and package dependencies. For more information,
read the [packages guide](https://docs.modular.com/max/packages.md).

### Does the MAX SDK collect telemetry?

Yes, the MAX SDK collects basic system information, session durations, compiler
events, and crash reports that enable us to identify, analyze, and prioritize
issues. The MAX container for model serving also collects performance metrics
such as time to first token and input processing time.

This telemetry is crucial to help us quickly identify problems and improve our
products for you. Without this telemetry, we would rely solely on
user-submitted bug reports, which are limited and would severely limit our
performance insights.

To disable serving telemetry, see the [MAX container
documentation](https://docs.modular.com/max/container.md#metrics).